Work as a Correlate of Canine Left Ventricular Oxygen Consumption, and the Problem of Catecholamine Oxygen Wasting

Transcription

1 273 Work as a Correlate of Canine Left Ventricular Oxygen Consumption, and the Problem of Catecholamine Oxygen Wasting G. ALEC ROOKE AND ERIC 0. FEIGL SUMMARY Neither stroke volume nor external cardiac work (the integral of pressure times flow during ejection) has been considered an important correlate of myocardial oxygen consumption. An initial set of experiments re-examined this question by independently varying heart rate, systolic blood pressure, and stroke volume in seven closed-chest, anesthetized dogs. This was achieved by cardiac pacing, a pressure control reservoir, phenylephrine infusion, and adjustment of arteriovenous shunts. Propranolol was used to minimize changes in contractility which might affect myocardial oxygen consumption. Stroke volume in the form of external work had a significant effect on oxygen consumption. From these results, a new pressure-work index of myocardial oxygen consumption was devised, and fitting parameters for the following indexes were determined: systolic pressure-rate product, estimated wall tension, external left ventricular work, triple product, mean pressure-rate product, E t (Bretschneider), and tension-time index. These indexes were prospectively applied to a second set of experiments in 11 closed-chest, anesthetized dogs given norepinephrine, isoproterenol, dobutamine, Nembutal, and propranolol to alter myocardial contractility. Inotropic oxygen wasting was observed with the tension-time, mean pressure-rate, triple product, and estimated wall tension indexes, but not with the pressure-work or systolic pressure-rate indexes. It is concluded that stroke work is an important correlate of myocardial oxygen consumption, and that the pressure-work or systolic pressure-rate indexes can account for catecholamine-induced changes in myocardial oxygen consumption without postulating an oxygen-wasting effect. Circ Res 50: , 1982 ONE of the earliest observations made on factors influencing the oxygen consumption of mammalian hearts was that the oxygen cost of "flow work" was less than that of "pressure work" (Evans and Matsuoka, 1914; Starling and Visscher, 1926). This led to the rejection of the idea that external stroke work (the integral of pressure times flow from the beginning to the end of ejection) is an important correlate of myocardial oxygen consumption. Oxygen consumption was considered to be related to indexes of total cardiac effort, such as the mean blood pressure times heart rate (Katz and Feinberg, 1958; Feinberg et al., 1962), estimated wall tension (Rodbard et al., 1964), and the tension-time index (Sarnoff et al., 1958). The latter was attractive because the pressure integral during systole appeared to account for the effect of changes in stroke volume as well as the effects of changes in blood pressure and heart rate on oxygen consumption. Similarly, the integral of wall tension (force) over time has been shown to account for the oxygen cost of shortening without requiring a term containing stroke volume (Weber and Janicki, 1977). In gen- From the Department of Physiology and Biophysics, University of Washington, Seattle, Washington. Supported by National Institutes of Health grants HL and GM Submitted in partial fulfillment for the requirements for a combined MD-PhD degree at the University of Washington. Address for reprints: Dr. Eric O. Feigl, Department of Physiology and Biophysics SJ-40, University of Washington, Seattle, Washington Received February 27,1981; accepted for publication October 15,1981. eral, stroke volume has not been found to be a useful predictor of oxygen consumption when combined with indexes that involve an integral over time. This is because, in the absence of contractility changes, large stroke volumes are associated with long systolic periods. On the other hand, stroke volume in the form of external work has continued to be useful when combined with indexes not containing a time integral. This has been most clearly demonstrated by the comparison of isovolumic and isotonic contractions when peak wall tension is kept constant. The additional oxygen consumed during isotonic contraction was correlated with external work (Britman and Levine, 1964; Burns and Covell, 1972). Despite the success of the combination of peak tension and external work, few data exist on the combination of the pressure-rate product and a term containing stroke volume. Multiple regression analysis of oxygen consumption against the mean pressure-rate product and cardiac output did not demonstrate a significant effect of cardiac output (Katz and Feinberg, 1958). A major problem of many indexes is the oxygenwasting effect of catecholamines. Simply stated, inotropic agents alter myocardial oxygen consumption more than is predicted by such indexes as the mean pressure-rate product or the tension-time index (Sarnoff et al., 1965; Sonnenblick et al., 1965; Chandler et al., 1968; Graham et al., 1968; Boerth et al., 1978). Bretschneider (1972) solved this prob-

2 274 CIRCULATION RESEARCH VOL. 50, No. 2, FEBRUARY 1982 lem by including an estimate of contractility (dp/ dt) in a predictive index. The goals of the present study were to reevaluate the importance of left ventricular stroke volume (and thus external cardiac work) on myocardial oxygen consumption, to devise a clinically useful index of myocardial oxygen consumption, and to reexamine the oxygen-wasting effect of catecholamines. In phase I of the investigation, heart rate, systolic blood pressure, and stroke volume were independently varied, while changes in myocardial contractility were minimized with propranolol. It was found that left ventricular stroke work is an important correlate of myocardial oxygen consumption and that the data could be fit with a new pressure-work index. In phase II, several indexes of myocardial oxygen consumption were prospectively applied to experiments in which heart rate, blood pressure, stroke volume, and contractility were altered with catecholamines. An oxygen-wasting effect was observed with the tension-time and other indexes, but not with the pressure-work or systolic pressure-rate indexes. Methods Preliminary Surgery Mixed breed dogs of either sex, weighing from 23 to 29 kg, had initial preparatory sterile surgery 8-23 days before the experiment. Following general anesthesia with Nembutal, 25 mg/kg, a right thoracotomy was performed, an electromagnetic flow probe (Zepeda Instruments) was implanted around the ascending aorta, and the leads brought out at the back of the neck. In phase I animals, but not in phase II, complete atrioventricular heart block was produced by cauterization of the bundle of His (Shiang et al., 1977). General Preparation Each dog received 2.5 mg/kg of morphine sulfate subcutaneously 1 hour prior to administration of 100 mg/kg of a-chloralose intravenously. Anesthesia was maintained with 10 mg/kg per hr continuous a-chloralose infusion plus 0.5-g supplements if needed. Positive pressure ventilation (Harvard 601 pump) with 5-cm H2O end-expiratory pressure was used to maintain end-expiratory carbon dioxide between 4% and 5%, as continuously monitored by an infrared absorption meter (Beckman LB-2). The inspired gas was enriched with oxygen to keep arterial oxygen tension between 125 and 150 mm Hg, as measured by an oxygen tension cuvette (Feigl and D'Alecy, 1971). The metabolic acidosis associated with chloralose anesthesia was counteracted by a continuous infusion of 75 mg NaHCO 3 / kg per hr (Arfors et al., 1971). Rectal temperature was held at 37 C by means of a temperature-controlled heating pad (Yellow Springs 73A). The animal was anticoagulated with a bolus injection of 750 U/kg heparin, iv, followed by a continuous infusion of 250 U/kg per hr. The experimental preparation is diagrammed in Figure 1. A double catheter-tip pressure transducer (Millar Instruments) was inserted in the left brachial artery and advanced until the distal transducer was in the left ventricle and the proximal transducer was in the ascending aorta. Coronary venous blood was continuously sampled with a Sones catheter (USCI #007538) inserted into the coronary sinus via the right jugular vein using a fluoroscope. The catheter tip was found to be mm inside the coronary sinus ostium on postmortem examination in the various experiments. Contamination of coronary sinus samples by right atrial blood was avoided by a low withdrawal rate of 10 ml/min using a roller pump (Koberstein et al., 1969). Aortic electromagnetic flowmeter L. vent, pressure - (shown inside LV) Syringe pump Pressure control reservoir ~ External jugular vein FIGURE 1 Diagram of the experimental preparation used to control stroke volume and systolic blood pressure independently, while measuring left ventricular oxygen consumption in closed-chest, anesthetized dogs. Stroke volume was primarily controlled with the arteriovenous shunts, whereas systolic pressure was primarily controlled with the pressure control reservoir and phenylephrine infusion. Not shown in the figure is a right ventricular pacing electrode for controlling the heart rate after surgical atrioventricular heart block.

3 CORRELATES OF MYOCARDIAL OXYGEN CONSUMPTION/Rooke and Feigl 275 Left circumflex coronary artery blood flow was measured with a cannula-tip ultrasonic Doppler shift flow transducer inserted via the right carotid artery in the closed-chest dog (Smith et al., 1974). At the end of the experiment, crystal violet dye was injected into the cannula, staining the perfused myocardium, which was cut out and weighed postmortem. This weight was used to express coronary flow in ml/min per 100 g of left ventricular tissue. In phase I experiments, but not in phase II, a bipolar pacing catheter was placed into the right ventricle via the right jugular vein. Heart rate was paced electrically with a stimulator synchronized to avoid stimulus artifacts on the electromagnetic flowmeter recording (Montgomery and Scher, 1974). Also in phase I experiments, but not in phase II, bilateral femoral arteriovenous shunts with screw clamps were used to vary left ventricular stroke volume. Hemodynamic variables (Fig. 2) were recorded on an oscillograph and also on a tape recorder for playback and computation. Oxygen Measurement Myocardial oxygen consumption was calculated as the product of coronary flow (ml/min per 100 g) and the arteriovenous oxygen difference. Coronary flow was electronically averaged (2-second time constant). Since arterial oxygen tension was maintained between 125 and 150 mm Hg, hemoglobin was assumed to be fully saturated and arterial oxygen content was calculated as the product of blood hemoglobin content as determined by the cyanmethemoglobin method (Bauer et al., 1974) and the constant 1.36 ml O 2 /g hemoglobin (Bernhart and ACFTiC f IC'N ( ers/tini intravt S'H =5: SS'.RE LEFT VENTRICULAR dfvdt 2.00 I I I 1.00 Ml AORTIC PRESSURE 'mm Hg} 75 0 C0R0NARY FLOW (ml/min per 100 g) -.TV t- '<- ' -i "-; - -( I! I I I '(-_ ',_ r<- 'f- 't- -",_ 'r- 'i- '""' r '-i '^t '"' '^ <v * ''-i "'i FIGURE 2 A record of the major hemodynamic variables studied in this investigation. Skeggs, 1943; Dijkhuizen et al., 1977). Oxyhemoglobin saturation of coronary sinus blood was measured continuously (0-600 Oximeter, Waters Instruments) during the experiment. A calibration curve for the oximeter was constructed by analyzing the hemoglobin and oxygen content (Lex-O 2 -Con, Lexington Instruments) (Kusumi et al., 1973) of coronary sinus samples during the experiment. Oxygen content of coronary sinus blood was calculated from the hemoglobin concentration and the oxyhemoglobin saturation during steady state. Left ventricular myocardial oxygen consumption was expressed as ml 0 2 /min per 100 g of perfused left ventricle. Experimental Design Phase I Stroke Volume Changes The objective of the experimental design in the first phase of this study was to examine the effect of changes in stroke volume on left ventricular myocardial oxygen consumption with fixed combinations of heart rate and systolic blood pressure. The data obtained in the first phase were used to estimate the parameters for several indexes of left ventricular oxygen consumption. Changes in contractility were minimized by administering propranolol, 1 mg/kg, iv, as a bolus, followed by 0.5 mg/kg per hr as a continuous iv infusion. Heart rate was set by right ventricular cardiac pacing following previous atrioventricular heart block. Stroke volume was determined primarily with adjustable arteriovenous shunts. Systolic blood pressure was experimentally controlled with a phenylephrine infusion and a pressure control reservoir connected to a femoral artery. The pressure control reservoir consisted of a collapsible plastic bag inside a box with air pressure in the box set to the desired level. If arterial pressure exceeded the box pressure, blood moved from the animal into the bag, and, conversely, from the bag to animal if the arterial pressure was below the box pressure. Manipulating shunts, phenylephrine infusion rate and the pressure control bottle interacted in their effects on stroke volume and systolic blood pressure so that a series of adjustments in diminishing steps was needed to achieve the desired stroke volume and systolic blood pressure. In phase I, all four combinations of two systolic blood pressures (120 and 180 mm Hg) and two heart rates (80 and 120 beats/min) were examined at two different stroke volumes, yielding eight data points per dog. The systolic blood pressure and heart rate values were chosen so that the middle products remained constant (120 mm Hg X 120 beats/min = 180 mm Hg x 80 beats/min). This permitted testing the linearity of a systolic blood pressure times heart rate term in predictive formulas. The stroke volumes had to be adjusted according to the cardiac ability of each dog, but the high stroke

4 276 CIRCULATION RESEARCH VOL. 50, No. 2, FEBRUARY 1982 volume was always chosen to be about 50% greater than the low stroke volume except at one of the 120 mm Hg systolic blood pressure, 120 beats/min heart rate points, where an intermediate stroke volume was chosen (Fig. 3). At the conclusion of each experiment, the stability of the preparation was tested by repeating one of the initial points. Since the original heart rate, systolic blood pressure, and stroke volume could not be duplicated exactly, the following correction was employed. The predicted oxygen consumption for both the original and duplicate points was calculated with the new pressure-work index and the difference between these two values was obtained. This small difference (typically 5%) was then added to the observed (measured) oxygen consumption of the duplicate point, and this adjusted value was compared with the observed oxygen consumption of the original point. The difference between these two oxygen consumptions ranged from 0.74 to ml 0 2 /min per 100 g. The mean of the differences equaled ± 0.52 (SD) ml 0 2 /min per 100 g, which was not significant by paired i-test stroke volume - a 32 ml A 23ml x 18 ml I dog 8 points X ONE DOG a 0 IOJOOO 20O00 SYSTOLIC PRESSURE - RATE PRODUCT (mm Hg/min) ED X X Phase II Contractility Changes The closed-chest preparation used in phase I was also used in phase II, except phase II dogs did not have atrioventricular heart block and cardiac pacing, and also did not have arteriovenous shunts or a pressure control reservoir. The intention was to prospectively test the formulas generated in the first phase during changes in myocardial contractility experimentally produced during the second phase. That is, the equations, including the fitting constants obtained by regression analysis on phase I data, were applied without change to the data from phase II. The hypothesis was that changes in left ventricular oxygen consumption due to alterations in contractility could be predicted from the phase I data base, where changes in contractility were small. The protocol consisted of infusing increasingly larger doses of inotropic drugs in half logarithmic increments (0.003, 0.01, etc.) until some pre-determined end point was reached. Isoproterenol was started at fig/kg per min continuous infusion and increased until heart rate exceeded 240, which usually occurred at 0.1 or 0.3 /xg/kg per min. Norepinephrine was given in the same doses, with the highest dose usually being 0.3 jug/kg per min. Here the end point was either a systolic pressure of 250 mm Hg or an increase in heart rate instead of a reflex decrease. Dobutamine was given in the dose range jug/kg per min, without a special end point. Each dog received a complete dose sequence of either isoproterenol or dobutamine. Ten of the 11 dogs also received a complete dose sequence of norepinephrine. When dobutamine was given, 0.1 /tg/kg per min isoproterenol was also given separately and later in combination with 10 mg/kg per min dobutamine. In addition, all but one dog received several doses of phenylephrine and myocardial depressants. Propranolol was administered as a 1-mg/kg single dose, followed by 0.5 mg/kg per hr, iv drip. Nembutal was given in a 1-mg/kg dose, followed by enough additional drug to total 3,10, or 30 mg/kg. I n SEVEN DOGS COMBINED BY POINT overoge stroke volume = a 32 ml a 26 ml x 20 ml 7 dogs 56 pants X 0 IOJOOO 20,000 SYSTOLIC PRESSURE - RATE PRODUCT (mm Hg/min) FIGURE 3 The increase in myocardial oxygen consumption resulting from an increase in stroke volume when the systolic pressure-rate product was held constant is illustrated. The results for an individual animal are shown in the left panel, while the data, averaged on a point-by-point basis, for all seven phase I experiments are shown in the right panel. Note that when stroke volume was held constant, there was a linear relation between the systolic pressure-rate product and myocardial oxygen consumption. The fixed increment in stroke volume required more oxygen consumption at a high systolic pressure-rate product than at a low value. In both panels, the systolic pressure-rate products were obtained with the following approximate values (SBP X HR): Left-most pair of points, 120 X 80 = 9,600. Middle four points, 120 X 120 = 14,400 or 180 x 80 = 14,400. Rightmost pair of points, 180 X 120 = 21,600. In the right panel, the standard error of the mean on the vertical axis ranged from 0.26 to 0.75 (ml 02/min per 100 g), which is approximately the height of the symbols. The standard error on the horizontal axis ranged from 120 to 203 (mm Hg x beats/min), which is a fraction of the width of the symbols. XX a X

5 CORRELATES OF MYOCARDIAL OXYGEN CONSUMPTION/.Roo&e and Feigl 277 Calculations A sample of an original record is shown in Figure 2. Average values were obtained in a 15- to 20- second interval during steady state over an integral number (2-4) of respiratory cycles. Heart rate (HR) was determined by dividing total beats by their time interval in minutes. Systolic blood pressure (SBP) in mm Hg was determined from the peak intraventicular pressure. Mean blood pressure in mm Hg was obtained by electronically averaging the aortic pressure signal (2-second time constant). Diastolic pressure (DBP) in mm Hg was obtained from the aortic pressure recording. Stroke volume (SV) in ml was obtained by integrating each beat of the aortic flow signal, employing the zero crossing of the flow signal to start and stop the integration (Integrator Coupler, Gould Inc.). The left ventricular pressure tension-time index integration was performed by gating the integrator with a square wave triggered from the electrocardiogram. A variable delay circuit triggered from the R wave started an adjustable duration square wave generator so that integration started at the onset of left ventricular systole and stopped at the end of systole. These timings were individually adjusted for each experimental run. As in previous studies, tension-time (mm Hg-sec) included both isovolumic periods, and end-diastolic pressure was not subtracted (Sonnenblick et al., 1965; Braunwald, 1971). Multiplication of the appropriate variables by heart rate yielded the systolic pressure-rate product, the mean pressure-rate product, and the tension-time index. The duration of ejection (t e jt) in seconds was determined from the aortic flow recording. This value, when multiplied by the systolic pressure-rate product, yielded the triple product. Neglecting left ventricular diastolic pressure, external left ventricular work is the integral of pressure times flow during ejection and was approximated by: (0.8 SBP DBP) x HR X SV. The diastolic pressure plus 0.8 of the pulse pressure has been shown to be a close estimate of the mean pressure during ejection (Walker et al., 1973). Stroke volume was normalized by dividing by the body weight (BW), as this has been shown to be an appropriate way to compare stroke volumes between different species and within a species (Holt et al., 1968). A formula for estimating peak wall tension from relatively easy-to-measure variables has been proposed by Rodbard et al. (1964). The formula is derived using a thin-walled, spherical model for wall tension and assumes a constant ejection fraction and the temporal association of peak wall tension with mean blood pressure. The formula is: mean BP x HR x SV 1/3, and is referred to as estimated wall tension. The maximum of the first derivative of pressure with respect to time, dp/dt, was obtained by differentiating the intraventricular pressure signal using a linear electronic differentiator with a 50-Hz cutoff. The units of dp/dt are meters of mercury per second (m Hg/sec). The duration of systole (ts VS ) was determined from the intraventricular pressure record during contraction and includes both isovolumic periods. Bretschneider's formula (Bailer et al., 1979) was calculated as follows: E,(ml 0 2 /min per 100 g) = Eo + Ei + E 2 + E 3 + E 4 where: Eo = 0.7 Ei = 0.03 x t sys X HR E 2 = ( X SBP 15 X tejt X HR)/(dP/dt) 1/3 E 3 = 1.2 x 10" 5 X dp/dt x HR E 4 = 8.0 x 10" 9 x (dp/dt) 15 x HR Data Analysis Indexes of myocardial oxygen consumption have usually been analyzed by retrospectively calculating a "best fit" slope and intercept plus a correlation coefficient that is used to judge the quality of the fit. In order to predict the oxygen consumption in ml 0 2 /min per 100 g, an entire equation including slope(s) and intercept must be defined. The units of the constants given in Table 1 are ml 0 2 /min per 100 g myocardium times the inverse of the hemodynamic units that the constant multiplies. The constant in a formula represents the average of the constants from regression analysis performed on each animal individually. Thus, each reported constant is the average of seven values from phase I data. No constants were developed for E t, since they were published previously (Bailer et al., 1979). The indexes analyzed in this paper were regressed on phase I oxygen consumption data three times once to determine a "best fit" intercept, once with a fixed intercept of 0.0, and once with a fixed intercept of Boerth et al. (1969) reported a myocardial oxygen consumption of 1.43 ml 0 2 /min per 100 g in the potassium-arrested heart a condition approximating the intercept where heart rate and systolic blood pressure are zero. Both the absolute (ml 0 2 /min per 100 g) and relative (percent change) predictive accuracy of the various oxygen consumption indexes were evaluated. The relative observed oxygen consumption for a given data point was calculated by dividing that oxygen consumption by the average of all the observed oxygen consumptions for that individual animal. The relative predicted oxygen consumption was calculated in a similar manner, using the corresponding predicted oxygen consumptions for an individual animal. This had the effect of comparing each dog to itself and eliminating the variability between dogs by making the mean observed and mean predicted values for each animal equal to 1.0 (Fig. 4). ; The predictive ability of the various indexes was evaluated by examining the error of prediction,

6 278 CIRCULATION RESEARCH TABLE 1 Phases I and II Stroke Volume and Inotropic Changes Phase I Stroke volume changes VOL. 50, No. 2, FEBRUARY 1982 Phase II Inotropic changes Absolute Relative Absolute Relative Formula Pressure-work index 7.36 x 10" 4 Syst pressure-rate x 10"" Syst pressure-rate x 10-" Syst pressure-rate X 10"' X Est. wall tension x 10" 3 x Est. wall tension x 10-' X Est. wall tension x 10" 5 External work x 10- r> External work X 10" 5 External work X 10" :1 Triple product X 10" 3 Triple product x 10" :l Triple product x 10" 4 Mean pressure-rate x 10" 4 Mean pressure-rate x 10~ 4 Mean pressure-rate E, (Bretschneider) Std. Std. Std. Std. Std. dp/dt Slope Std dp/dt Slope % Dev. Dev. Corr. Dev. Corr. Dev. Dev. Corr. I ml O 2 \ rj ev Corr. / \ (ml O 2 ) (%) Coeff. (%) Coeff. (ml O 2 ) (%) Coeff. \m Hg/sec/ (%) Coeff. \m Hg/sec/ * * * * * * * * * * * * * * * * * * * * * * * * * 3.86 x 10~ 3 Tension-time index X 10" 3 Tension-time index x 10" 3 Tension-time index * * * * * * The first formula in each group of three is the regression equation when both slope and intercept were best fit. The second formula is when the intercept was fixed at zero, and the third is when the intercept was fixed at * Slope significantly different from zero (P < 0.01).

7 CORRELATES OF MYOCARDIAL OXYGEN CONSUMPTION/Rooke and Feigl 279 PRESSURE-WORK INDEX TENSION-TIME INDEX 7 dogs 56 points 7 dogs 56 points RELATIVE PREDICTED OXYGEN CONSUMPTION (ratio).7 1X RELATIVE PREDICTED OXYGEN CONSUMPTION (ratio) FIGURE 4 A comparison of the pressure-work and tension-time indexes when applied retrospectively to phase I (stroke volume changes) relative oxygen consumption data. The pressure-work index was superior to the tension-time index when between-dog variability was removed by comparing each dog to itself. These data have been normalized to remove between-dog variability by comparing each dog to itself. Each data point is expressed as the ratio of that value to the average value (1.0) of all the points for that individual dog. A fixed intercept of 1.43 was used with each index. Each symbol represents a different dog. The line of unity represents perfect prediction. which is the observed oxygen consumption minus the predicted oxygen consumption for any given data point. The standard deviation of the error of prediction was calculated as: Std. dev. = ( (observed - predicted ) 2 /n) 1/2 where n is the number of points. This calculation differs from the conventional standard deviation in that the mean of the differences is not subtracted from the square of each difference. If a regression formula were applied to the same data from which it was developed, the mean of the differences would equal zero and the two standard deviations would be identical (except for the divisor, n vs. n 1). When a regression formula is applied to new data, the prediction may be consistently too high or too low, and this error should not be neglected. The standard deviations in the figures and table are given in experimental units (ml 02/min per 100 g) or in percent of the mean myocardial oxygen consumption. The mean oxygen consumption in phase I was ml 0 2 /min per 100 g and in phase II was ml 0 2 /min per 100 g. In phase I, the effect of stroke volume on myocardial oxygen consumption and the presence of an interaction between stroke volume and the systolic pressure-rate product were tested by a 2 X 3 factorial analysis of a randomized complete block design (analysis of variance). Each dog represented one block (seven blocks total). The two factors were stroke volume at two levels (low = 20 ml and high = 32 ml) and the systolic pressure-rate product at three levels (low = 9,600, medium = 14,400, and high = 21,600). The factorial analysis provided a statistical test of the overall significance of the (main) effect of stroke volume on oxygen consumption and also tested the interaction of stroke volume and systolic pressure-rate product on oxygen consumption (Ostle, 1954). In phase II, the relationship of the prediction errors to the contractile state of the heart was determined by regressing the errors against dp/dt. If the slope of the regression equation was a significantly positive value, then inotropic "oxygen wasting" was demonstrated. When considering absolute oxygen consumption, the slope has the units of ml 0 2 /min per 100 g per m Hg/sec. For example, if the dp/dt slope constant was 1.0, then a change in dp/ dt of 5 m Hg/sec predicted an underestimation of 5 ml 0 2 /min per 100 g. When considering relative oxygen consumption, the slope has the units of percent relative error per m Hg/sec. For example, a slope of 10.0 combined with a change in dp/dt of 5 m Hg/sec predicted an underestimation of 50% of the mean relative oxygen consumption. Standard regression analysis was performed using SPSS, a preprogrammed computer package (Nie et al., 1975). Results Phase I When systolic blood pressure and heart rate were held constant, an increase in stroke volume was accompanied by an increase in left ventricular oxygen consumption, as illustrated in Figure 3. This was observed in all 28 comparisons in seven dogs. The effect was significant by factorial analysis of variance (P < 0.001). Furthermore, it can be observed in Figure 3 that a fixed increment in stroke volume cost more oxygen when the systolic blood pressure was high (180 mm Hg) than low (120 mm Hg) (P < by analysis of variance), implying that external stroke work (stroke volume X mean pressure during ejection) is more important than stroke volume alone. This interaction may also be stated conversely, in that a given increment in the systolic pressurerate product with a constant stroke volume resulted in greater oxygen consumption when stroke volume

8 280 CIRCULATION RESEARCH VOL. 50, No. 2, FEBRUARY 1982 was fixed at a high level than when fixed at a low level. That is, the slope of the high stroke volume points ( ) is greater than the low stroke volume points (X) in Figure 3. The interaction between stroke volume and systolic blood pressure is unlikely to have been due to reflex changes in contractility, since myocardial /?-adrenergic receptors were blocked with propranolol. Examination of points of constant stroke volume in Figure 3 shows the systolic pressure-rate product to be related linearly to oxygen consumption. In addition, proportional changes in heart rate and systolic pressure appear to have equal effects on oxygen consumption. This is indicated by the two middle X data points in either panel of Figure 3, which have nearly equal oxygen consumptions at similar systolic pressure-rate products, despite the fact that the cross-products are achieved by different combinations of pressure and rate. In one case, the systolic blood pressure was approximately 180 mm Hg and the heart rate 80 beats/min (180 X 80 = 14,400); in the other case, systolic blood pressure was approximately 120 mm Hg and heart rate 120 beats/min (120 X 120 = 14,400). This observation was statistically confirmed by multiple regression performed on the logarithm of relative observed oxygen consumption above the arrested-heart value of 1.43 (1.43 was subtracted from the observed oxygen consumption, and the remainder was expressed as a percentage of the average of all values for that dog). The independent variables were the log of systolic pressure and the log of heart rate. When regression was performed on points of constant stroke volume, systolic pressure was found to have an exponent of 1.06 and a 95% confidence interval of 0.94 to 1.18, whereas heart rate had an exponent of 0.96 and a 95% confidence interval of 0.86 to The data were fit by an equation that combined systolic pressure-rate product and left ventricular external work divided by body weight (see Methods). The new formula, named the pressure-work index, is: MV0 2 = K,(SBP x HR) + K 2 where: (0.8 SBP DBP) x HR x SV> BW MV O 2 = left ventricular myocardial oxygen consumption (ml (Vmin/lOO g) SBP = systolic blood pressure (mm Hg) DBP = diastolic blood pressure (mm Hg) HR = heart rate (beats/min) SV = stroke volume (ml) BW = body weight (kg) K, = 4.08 X 10~ 4 K 2 = 3.25 X 10" 4 The results of applying the pressure-work index retrospectively to phase I data and a comparison to the tension-time index are shown in Figure 4. Between-dog variability was removed by comparing each animal to itself. The pressure-work index accounts for the effect of stroke volume on myocardial oxygen consumption. The complete formulas and their statistics are given in Table 1. The phase I results are summarized in Figure 5. The difference between relative and absolute oxygen consumption predictions in Figure 5 indicates the magnitude of the betweendog variability for a formula, except for E t. The standard deviation for absolute oxygen consumption was large for E t, because previously published fitting constants were used prospectively. The error for relative oxygen consumption using E t may be appropriately compared to the errors of the other formulas. Phase II The indexes developed in phase I, when /?-adrenergic receptor blockade prevented large changes in contractility, were prospectively applied to phase II experiments where contractility was altered by infusion of norepinephrine, isoproterenol, dobutamine, Nembutal, and propranolol. The average of the lowest value of peak dp/dt for each of the seven dogs in phase I was 1.73 m Hg/sec and the average highest value was 3.58 m Hg/sec (difference = 1.85 m Hg/sec). In phase II, the average low value in 11 dogs was 1.41 m Hg/sec and the average high value was 8.46 m Hg/sec (difference = 7.05 m Hg/sec). Pressure - Work Index E t (Bretschneider) Triple Product Tension-Time Index Estimated Wall Tension Systolic Pressure-Rate External Work Mean Pressure-Rate 14.4 = Relative = Absolute PERCENT FIGURE 5 Percent standard deviations of the error in prediction (observed-predicted) of absolute and relative oxygen consumptions for phase I (stroke volume changes) data analyzed retrospectively. A fixed intercept of 1.43 was used (except EJ.

9 CORRELATES OF MYOCARDIAL OXYGEN CONSUMPTION/i?oo>fee and Feigl 281 PRESSURE-WORK INDEX TENSION- TIME INDEX S. D. aboui line of unity = 13.7%? > 30 S.D. about line of unity = 34.4% line of unity dogs 231 points 11 dogs 231 points PREDICTED OXYGEN CONSUMPTION PREDICTED OXYGEN CONSUMPTION (ml 0 2/min per loog) (ml 0 2/min per loog) FIGURE 6 A comparison of the pressure-work and tension-time indexes when applied prospectively to phase II (inotropic changes) absolute oxygen consumption data. The pressure-work index was superior to the tension-time index. The indexes have intercepts of The line of unity represents perfect prediction. The pressure-work index and the tension-time index are compared in Figure 6, with respect to the ability to predict absolute oxygen consumption during changes in contractility. The superiority of the pressure-work index is evident, and this is even more apparent if the same data are expressed as relative oxygen consumption where each dog is compared to itself, as shown in Figure 7. The question of "inotropic oxygen wasting" may be approached by examining the prediction error (the difference between observed and predicted oxygen consumption) as a function of contractility, as estimated by peak dp/dt. If a formula accurately predicts myocardial oxygen consumption during changes in contractility, there will be a small error, PRESSURE-WORK INDEX and this error will not be correlated with dp/dt. If inotropic oxygen wasting occurs, then a significant positive correlation between the prediction error and dp/dt will be observed, such that the observed oxygen consumption at high levels of contractility will be greater than predicted. These effects may be observed in Figures 8 and 9, where no inotropic oxygen wasting is demonstrated with the pressurework index but a large, significant oxygen-wasting effect is demonstrated with the tension-time index. The predictive abilities of the various indexes when prospectively applied to the phase I data are summarized in Figure 10 and Table 1. Errors in prediction were found to have a significant positive correlation with dp/dt in the formulas containing the TENSION- TIME INDEX RELATIVE PREDICTED OXYGEN CONSUMPTION (ratio).4 RELATIVE PREDICTED OXYGEN CONSUMPTION FIGURE 7 A comparison of the pressure-work and tension-time indexes when applied prospectively to phase II (inotropic changes) relative oxygen consumption data. These data have been normalized to remove between-dog variability by comparing each dog to itself. Each data point is expressed as the ratio of that value to the average value (1.0) of all the points for that dog. The pressure-work index was much superior to the tension-time index. The indexes have intercepts of The line of unity represents perfect prediction. (ratio)

10 282 CIRCULATION RESEARCH VOL. 50, No. 2, FEBRUARY PRESSURE-WORK INDEX TENSION-TIME INDEX 16 regression line 0 ^^ y~ v^ regression line.}.-} * * ; \ '" v y ^line of no error * line of no error' r dogs P = points p = points MAXIMUM dp/dt -8 r =.839 MAXIMUM dp/dt (m Hg/sec) (m Hg/sec) 11 dogs FIGURE 8 The errors in prediction (absolute observed-absolute predicted; that is, the vertical distance a point lies from the line of identity in Fig. 6) vs. peak dp/dt are plotted for the pressure-work and tension-time indexes using phase II (inotropic changes) data. These unnormalized data contain between-dog variability, as well as the effect of altered contractility. No trend (correlation) was observed between the error in prediction and contractility for the pressure-work index, but a large, significant correlation was observed with the tension-time index. The significant underestimation ofmyocardial oxygen consumption with high contractility is the inotropic oxygen wasting effect found with the tension-time index. The line of no error represents perfect prediction. triple product and the mean pressure-rate product, but not when the pressure-work index or the systolic pressure-rate product were applied. The choice of intercept modestly affected the magnitude of the error vs. dp/dt correlation because oxygen consumption was correlated with dp/dt. An intercept u r =.004 p=.48 that was too low created positive errors at low oxygen consumptions which resulted in a more negative error vs. dp/dt slope. This was best illustrated in the case of external work. The Bretschneider index, E t, did poorly in phase II because it no longer had a linear relationship to oxygen consumption. PRESSURE-WORK INDEX TENSION-TIME INDEX regression line + line of no error ^ ' ' regression line dogs points r =.905 p = dogs points MAXIMUM dp/dt MAXIMUM dp/dt (m Hg/sec) (m Hg/sec) FIGURE 9 The errors in prediction (relative observed-relative predicted; that is, the vertical distance a point lies from the line of identity in Fig. 7) vs. peak dp/dt are plotted for the pressure-work and tension-time indexes using phase II (inotropic changes) data. These data have been normalized to remove between-dog variability by comparing each dog to itself. Each data point is expressed as the ratio of that value to the average value (1.0) of all the points for that dog. No trend (correlation) was observed between the error in prediction and contractility for the pressure-work index, but a large, significant correlation was observed with the tension-time index. The significant underestimation ofmyocardial oxygen consumption with high contractility is the inotropic oxygen wasting effect found with the tensiontime index. The line of no error represents perfect prediction.

11 CORRELATES OF MYOCARDIAL OXYGEN CONSUMPTION/Rooke and Feigl 283 Pressure-Work Index Systolic Pressure-Rate External Work Estimated Wall Tension Triple Product Mean Pressure-Rate E, (Bretschneider) Tension-Time I I = Relative = Absolute Index '/////////////A ////////////////A /////////////I YYYYYYYYYYYYYYA 24.7 ////////////////I V//////////////77//7A Y///Y////////////7M PERCENT FIGURE 10 Percent standard deviations of the error in prediction (observed-predicted) of absolute and relative oxygen consumptions for phase II (inotropic changes) data analyzed prospectively. A fixed intercept of 1.43 was used (except E t ). Discussion The goals of the present study were to reevaluate the importance of external left ventricular work as a correlate of myocardial oxygen consumption, devise a clinically useful index of left ventricular oxygen consumption, and reexamine the oxygen-wasting effect of catecholamines in a closed-chest preparation. The utility of an index is related to the ease that it may be employed in a clinical setting. For this reason we deliberately limited the measured variables to heart rate, blood pressure and cardiac output variables that may reasonably be obtained in patients. Undoubtedly, better indexes (correlations) may be obtained if more variables, such as ventricular volume and left ventricular dp/dt, are employed. However, at some point it makes better sense to measure myocardial oxygen consumption directly than to improve the index with additional hemodynamic variables. The results of phase I demonstrate that stroke volume has a significant effect on left ventricular oxygen consumption, and that systolic pressure, heart rate, and stroke volume all interact in their effects on oxygen consumption. Independent experimental control of these three variables in each phase I dog made it possible to demonstrate a role for stroke volume and/or external left ventricular work. It is likely that the variability between dogs has obscured this observation in some previous studies. For example, when the aggregated data from phase I were regressed against systolic pressure-rate product and stroke volume, a correlation with stroke volume was just barely significant by multiple regression analysis. Yet the importance of stroke volume was apparent in each animal when high and low stroke volume comparisons were made at matched systolic pressure-rate products (Fig. 3). The units of energy and work are identical (e.g., joules), and it makes thermodynamic sense to characterize the energetics of any engine in terms of the work it does. Early studies (Evans and Matsuoka, 1914; Starling and Visscher, 1926) indicated that external cardiac work alone was poorly correlated with myocardial oxygen consumption, and this led to the incorporation of pressure in a number of indexes. Whereas pressure development is clearly well correlated with myocardial oxygen consumption, it seems unlikely that cardiac work is entirely without metabolic cost. The pressure-work index developed here combines these two factors in a clinically applicable index. It should be emphasized that all indexes are only correlations and that the data may be satisfactorily fit by a number of indexes, as indicated in Table 1. If left ventricular volume had been measured and wall tension calculated, it is likely that a very satisfactory correlation between developed wall tension and myocardial oxygen consumption could have been made. On the other hand, previous work clearly indicates that there is a Fenn effect for cardiac muscle (additional oxygen consumption associated with external work) (Britman and Levine, 1964; Coleman et al., 1969; Burns and Covell, 1972), suggesting that external work is a sensible correlate of myocardial oxygen consumption. The relative influence of the systolic pressurerate product and external cardiac work may be evaluated by calculating a solution to the pressurework index. A 25-kg dog with a blood pressure of 120/80 mm Hg, a heart rate of 80 beats/min, and a stroke volume of 25 ml would have a predicted oxygen consumption of 8.26 ml 02/min per 100 g. Of this, 48% would be accounted for in the systolic pressure-rate product term, 35% in the external work term and 17% in the arrested-heart value of Under different hemodynamic conditions, the proportions would be somewhat different, but it is interesting to note that about 40% of the oxygen consumption above that found in the arrested condition is associated with the external cardiac work term. Previous studies have demonstrated that there is not an exact proportion between heart rate and myocardial oxygen consumption. When pressure and cardiac output were held constant, increasing heart rate decreased stroke volume and a doubling of heart rate resulted in augmenting oxygen consumption by only about 50% (Berglund et al., 1958; Laurent et al., 1956; Kralios et al., 1978), while, if stroke volume was held constant (increased cardiac output) during tachycardia, a greater augmentation

12 284 CIRCULATION RESEARCH VOL. 50, No. 2, FEBRUARY 1982 in myocardial oxygen was observed (Kralios et al., 1978). The present study confirmed the preceding studies and demonstrated that changes in heart rate resulted in proportionally equal changes in extrabasal oxygen consumption, provided that stroke volume remained constant. Other formulas have combined external cardiac work with a term representing the energy required to generate pressure in the ventricle. The first of these was the contractile element work index of Britman and Levine (1964) that included an internal work term proportional to peak wall tension. Internal work against an elastic element and external work were given equal weighting. Later investigations demonstrated that external work cost less oxygen than internal work (Pool et al., 1968; Coleman et al., 1969; Burns and Covell, 1972). An interesting new approach is to measure the energy remaining in the ventricle at the end of ejection (but before isovolumic relaxation) and combine that energy with external work (Khalafbeigui et al., 1979; Suga, 1979). Indexes of myocardial oxygen consumption that do not include stroke volume or external cardiac work often give good results (Katz and Feinberg, 1958; Sarnoff et al, 1958; Feinberg et al., 1962; Bretschneider, 1972; Weber and Janicki, 1977; Bailer et al., 1979). All of these indexes that could be applied to the closed-chest dog without measuring left ventricular volume have been tested in the present investigation. As summarized in Figures 5 and 10, each of these indexes had greater prediction errors than the pressure-work index under the conditions of changing stroke volume and/or contractility. Propranolol was used in phase I to block /S-receptor-mediated changes in contractility so that the fitting constants for the various indexes could be determined without catecholamine effects. Contractility was probably not completely constant in phase I because of the interval-strength (Treppe) effect of heart rate on contractility. However, the contractility change associated with changing the heart rate from 80 to 120 is not large in canine hearts (Mitchell et al, 1963; Noble et al, 1969). The idea of catecholamine oxygen wasting is predicated on knowing what the oxygen consumption ought to be by some formula (index) without directly measuring myocardial oxygen consumption. That is, a predictive index of myocardial oxygen consumption is a prerequisite to the concept of oxygen wasting. An essential point in the design of the present study was that several indexes of myocardial oxygen consumption were developed from phase I data where there was little change in contractility, and that these indexes were then prospectively applied (without change) to phase II data where contractility was experimentally altered. The logic was that if inotropic oxygen wasting occurs, it should be demonstrated with an index developed from data without inotropic changes. As detailed in Table 1, estimated wall tension, the triple product, mean pressure-rate product, and especially the tension-time index, showed inotropic oxygen wasting (significant positive slopes vs. dp/dt), whereas the pressure-work index and systolic pressure-rate product index did not. One problem with oxygen wasting is that the magnitude depends on the formula used (Table 1). External work and the Bretschneider index showed significant negative slopes vs. dp/dt. This would indicate an oxygen-sparing effect of catecholamines, and emphasizes the point that oxygen wasting is dependent on the index used to demonstrate it. If the tension-time index is used as an index of myocardial oxygen consumption, then a large, and significant (P < ), inotropic oxygen-wasting effect has been demonstrated by these data. The tension-time index is particularly susceptible to showing oxygen wasting because the duration of systole becomes briefer with increasing contractility and heart rate. On the other hand, if the pressurework index or systolic pressure-rate product is used, then there is no evidence for inotropic oxygen wasting. Since the concept of catecholamine oxygen wasting requires an index, these data demonstrate that oxygen wasting is an arbitrary result of the index that is employed. When developed tension was used as an index of myocardial oxygen consumption, the additional oxygen consumption associated with catecholamine infusion could be accounted for by a velocity of contraction term (Coleman et al, 1971). Inspection of Figure 9 indicates that a similar accounting for catecholamine effects could be made with the tension-time index by retrospectively fitting an additional dp/dt term to the phase II data. The present experiments indicate that when indexes and fitting constants are determined in the absence of /?-adrenergic receptor-mediated contractility changes (phase I with propranolol), further adjustment and contractility-dependent terms do not need to be added to the pressure-work or systolic pressure-rate indexes to account for catecholamine effects (phase II). This indicates that when the same level of cardiac performance is achieved with catecholamines or by changes in preload and afterload, the same oxygen consumption results, without an additional oxygen cost due to a special effect of catecholamines. This result should not be over-interpreted; if the heart rate, systolic blood pressure, and cardiac work are augmented by catecholamines, this clearly increases left ventricular oxygen consumption, but not because of an oxygen-wasting effect of catecholamines. If cardiac output cannot be measured or estimated, then the data indicate that the systolic pressure-rate product with the following constants is likely to be the most useful index of left ventricular oxygen consumption. MV0 2 = 7.20 X 10" 4 (SBP X HR)

13 CORRELATES OF MYOCARDIAL OXYGEN CONSUMPTION/Rooke and Feigl 285 Bailer and co-workers (1981) also observed that systolic pressure-rate product is a good index of oxygen consumption over a wide range of values. As indicated in Figures 5 and 10, all indexes have smaller prediction errors if the percent (relative) change in myocardial oxygen consumption for an individual dog is calculated, rather than the absolute oxygen consumption. This is because betweensubject variability is added to the estimate when an absolute (ml 02/min per 100 g) value is called for. Thus relative changes in an individual's myocardial oxygen consumption may be calculated with less error than absolute values. The prediction of human myocardial oxygen consumption will be improved by determining the constants for the formula in humans. There may be problems in estimating left ventricular systolic pressure from the peripheral pulse because of normal pulse wave amplification or pathological aortic valve stenosis. It is likely that a toxic level of catecholamine infusion will show oxygen wasting even with the pressure-work index. The indexes developed in this study need to be tested during exercise, myocardial hypertrophy, and the like. The application of these indexes in heart failure is especially questionable without experimental verification. In conclusion, stroke volume in the form of external work was found to be an important correlate of left ventricular myocardial oxygen consumption in an experiment where systolic pressure, heart rate, and stroke volume were independently varied. A pressure-work index was developed to account for these variables and their interactions on myocardial oxygen consumption. Inotropic oxygen wasting was not observed using the pressure-work or systolic pressure-rate indexes. Acknowledgments We thank Stephanie Lathrop and Fellner Smith for expert technical assistance in all phases of this study. References Arfors K-E, Arturson G, Malmberg P (1971) Effect of prolonged chloralose anesthesia on acid-base balance and cardiovascular functions in dogs. Acta Physiol Scand 81: Bailer D, Bretschneider HJ, Hellige G (1979) Validity of myocardial oxygen consumption parameters. Clin Cardiol 2: Bailer D, Bretschneider HJ, Hellige G (1981) A critical look at currently used indirect indices of myocardial oxygen consumption. Basic Res Cardiol 76: Bauer JD, Ackermann PJ, Toro G (1974) Clinical Laboratory Methods, ed 8. St. Louis, CV Mosby, pp Berglund E, Borst HG, Duff F, Schreiner GL (1958) Effect of heart rate on cardiac work, myocardial oxygen consumption and coronary blood flow in the dog. Acta Physiol Scand 42: Bernhart FW, Skeggs L (1943) The iron content of crystalline human hemoglobin. J Biol Chem 147: Boerth RC, Covell JW, Pool PE, Ross J Jr (1969) Increased myocardial oxygen consumption and contractile state associated with increased heart rate in dogs. Circ Res 24: Boerth RC, Hammermeister KE, Warbasse JR (1978) Comparative influence of ouabain, norepinephrine and heart rate on myocardial oxygen consumption and inotropic state in dogs. Am Heart J 96: Braunwald E (1971) Control of myocardial oxygen consumption. Physiologic and clinical considerations. Am J Cardiol 27: Bretschneider HJ (1972) Die Hamodynamischen Determinanten des myokardialen Sauerstoffverbrauches. In Die Therapeutische Anwendung /S-sympathikolytischer Stoffe. edited by HJ Dengler. Stuttgart-New York, Schattauer, pp Britman NA, Levine HJ (1964) Contractile element work: a major determinant of myocardial oxygen consumption. J Clin Invest 43: Burns JW, Covell JW (1972) Myocardial oxygen consumption during isotonic and isovolumic contractions in the intact heart. Am J Physiol 223: Chandler BM, Sonnenblick EH, Pool PE (1968) Mechanochemistry of cardiac muscle. III. Effects of norepinephrine on the utilization of high-energy phosphates. Circ Res 22: Coleman HN, Sonnenblick EH, Braunwald E (1969) Myocardial oxygen consumption associated with external work: The Fenn effect. Am J Physiol 217: Coleman HN, Sonnenblick EH, Braunwald E (1971) Mechanism of norepinephrine-induced stimulation of myocardial oxygen consumption. Am J Physiol 221: Dijkhuizen P, Buursma A, FongersTME, Gerding AM, Oeseburg B, Zijlstra WG (1977) The oxygen binding capacity of human haemoglobin. Pfluegers Arch 369: Evans CL, Matsuoka Y ( ) The effect of various mechanical conditions on the gaseous metabolism and efficiency of the mammalian heart. J Physiol (Lond) 49: Feigl EO, D'Alecy LG (1971) Cuvette for continuously measuring blood oxygen tension. J Appl Physiol 30: Feinberg H, Katz LN, Boyd E (1962) Determinants of coronary flow and myocardial oxygen consumption. Am J Physiol 202: Graham TP, Covell JW, Sonnenblick EH, Ross J Jr, Braunwald E (1968) Control of myocardial oxygen consumption: Relative influence of contractile state and tension development. J Clin Invest 47: Holt JP, Rohde EA, Kines H (1968) Ventricular volumes and body weight in mammals. Am J Physiol 215: Katz LN, Feinberg H (1958) The relation of cardiac effort to myocardial oxygen consumption and coronary flow. Circ Res 6: Khalafbeigui F, Suga H, Sagawa K (1979) Left ventricular systolic pressure-volume area correlates with oxygen consumption. Am J Physiol 237: H566-H569 Koberstein RC, Pittman DE, Klocke FJ (1969) Right atrial admixture in coronary venous blood. Am J Physiol 216: Kralios AC, Tsagaris TJ, Kuida H (1978) Comparison of rhythm and rate-induced myocardial O2 consumption changes before and after coronary occlusion. Am Heart J 96: Kusumi F, Butts WC, Ruff WL (1973) Superior analytical performance by electrolytic cell analysis of blood oxygen content. J Appl Physiol 35: Laurent D, Bolene-Williams C, Williams FL, Katz LN (1956) Effects of heart rate on coronary flow and cardiac oxygen consumption. Am J Physiol 185: Mitchell JH, Wallace AG, Skinner NS Jr (1963) Intrinsic effects of heart rate on left ventricular performance. Am J Physiol 205: Montgomery GD, Scher AM (1974) A stimulator for use with electromagnetic flowmeters. J Appl Physiol 37: Nie NH, Hull CH, Jenkins JG, Steinbrenner K, Bent DH (1975) Statistical Package for the Social Sciences, ed 2. New York, McGraw-Hill Noble MIM, Wyler J, Milne ENC, Trenchard D, Guz A (1969) Effect of changes in heart rate on left ventricular performance in conscious dogs. Circ Res 24: Ostle B (1954) Statistics in Research. Ames, Iowa, Iowa State College Press, pp Pool PE, Chandler BM, Seagren SC, Sonnenblick EH (1968) Mechanochemistry of cardiac muscle. II. The isotonic con-