Correspondence OXYGEN CONSUMPTION IS INDEPENDENT OF CHANGES IN OXYGEN DELIVERY IN SEVERE ADULT RESPIRAlORY DISTRESS SYNDROME
tween O, delivery and O, consumption in the adult respiratory distress syndrome. Chest 1983; 81:267-71. 3. Kariman K, Buins 8R. Regulationof tissue oxygenextraction is disturbed in adult respiratory distress syndrome. Am Rev Respir Dis 1985;132:109-14.
To the Editor: Ronco and coworkers, (American Review of Respiratory Disease 1991; 143:1267-1273) concluded in their paper "Oxygen Consumption Is Independent of Changes in Oxygen Delivery in Severe Adult Respiratory Distress Syndrome" that the finding of pathologic dependence of oxygen consumption on oxygen delivery in ARDS in previous studies (1, 2, 3) may be due to methodological error. We have the following fundamental problem with their study: (1) The authors used an argument that cardiac output as measured by thermodilution has a systematic overestimation of the true cardiac output and, therefore, explained the apparent mathematical coupling of calculated VO:z and Oo:z. However, they then argued that the random variation in thermodilution cardiac output and the respective 00:z was less pronounced than those calculated by Fick technique [Le., 00:z = Cao, (Vo:z/Cao:z - Cvo:z)]. Therefore, the thermodilution-derived i>o:z should be assumed to be the more precise values. Patients 1 and 2 in their study had an approximately 24 and 17070 increase in oxygen delivery respectively and demonstrated no change in the calculated VO:z or in measured Vo:z. Therefore, there was no evidence of oxygen supply dependency in these two patients. The question arises that if the mathematical coupling plays a major role in oxygen supply dependency calculation, why didn't these two patients exhibit any increase in the calculated VO:z? These two patients, plus previous studies (1, 3) confirm that there are patients who exhibit no oxygen supply dependency (with either method), despite the significant increases in oxygen delivery. Does mathematical coupling discriminate against particular patients? (2) Ronco and colleagues indicated the excellent precision of the measured VO:z (1.8070), Caos, and Cvo, in their paper. However, if the variation in these parameters is random, why was a systematic reduction of 13070 in the Fick-derived cardiac output observed (from 6.9 l/min to 6.0 IImin) after blood transfusion? Although Ronco and coworkers try to address this issue in their discussion, it is very apparent that they failed to convince at least this reader that the variation of thermodilution in cardiac output will be any less than the Fick cardiac output. Even if we accept that the random variation in measurement of Cao, and Cvo, and measured oxygen consumption were in the same direction, which statistically will be extremely unlikely, one would have difficulty explaining the 13070 reduction in the derived cardiac output and no change in oxygen delivery. Therefore, we do not accept their speculation that the finding of pathologic dependence of oxygen consumption on oxygen delivery in patients with ARDS in previous studies was due to only methodological error. ZAB MOHSENIFAR, M.D. Director Respiratory Therapy Associate Professor of Medicine DAVID J. Ross, M.D. Medical Director Lung Transplantation Assistant Professor of Medicine Cedars-Sinai Medical Center Los Angeles, CA 1. Danek 81, Lynch IP, Weg10, Dantzker DR. The dependence of oxygen uptake on oxygen delivery in the adult respiratory distress syndrome. Am Rev Respir Dis 1980; 122:387-95. 2. Mohsenifar Z, Ooldbach P, Tashkin DP, Campisi Dl, Relationship be978
From the Authors: We interpret the concerns of Drs. Mohsenifar and Ross as, "In evaluating the relationship between oxygen consumption and delivery, how do we justify use of oxygen consumption measured using a metabolic cart (Vo:zcart) while using a different methodology for oxygen delivery based on thermodilution cardiac output (Dostherm)?" We justify our choice of using apparently contrasting methodologies based on considerations of accuracy of the measurement, power to detect a change in the measurement, and by examining potential sources of measurement error for the two methods of determining VO:z, Le.,Vo:zcart compared with Vo:ztherm determined from calculations based on thermodilution cardiac output, and for the two methods of determining Dos, i.e., Dostherm compared with Do-cart determined from calculations based on Fick cardiac output. As we will show, Vo:zcart is preferred over Vo:ztherm because of greater accuracy and power to detect changes in VO:z and is not subject to error from mathematical coupling related to change in hemoglobin after blood transfusion. In contrast, Dostherm is preferred over Do-cart because of considerations of power to detect changes in Do.. Further to preference of Vo:zcart and Dostherm based on considerations of accuracy and power to detect change is that the relationship between Vo:zcart and Dostherm is not compromised by error from mathematical coupling. We believe that we are justified in concluding that VO:z did not change after a significant change in Do, after blood transfusion. We first assess accuracy and power to detect change of the two methodologies for VO:z. We performed validation of Vo:zcart over a range of ventilator settings used for patients who have adult respiratory distress syndrome (ARDS) using an in vitro system whereby we simulated known VO:z using dilution of F'Io:z with precisely regulated flow of N:z (J Crit Care 1991, 6:36). Vo:zcart had an overall measurement error of 1.8070 up to F'Io:z 0.80, minute ventilation 20 L/min, and peak inspiratory pressures of 80 em H:zO. Coefficient of variation of in vitro measurement of Vo:zcart in patients who had ARDS ranged from 4070 at F'Io:z < 0.55 to 7070 at F'Io:z 0.80. We did not perform tests of accuracy of Vo:ztherm. However, accuracy of thermodilution cardiac output is 4 to 10070 as reviewed by others (Am Rev Resp Dis 1982; 126:1001). Using principles for calculating the variances of a product of independent variables (Introduction to the theory of statistics, 3rd edition, Toronto: McGraw-Hill, 1974; 180), we estimate coefficient of variation for Vo:ztherm to be 8070 based on means of our measurements of coefficients of variation for components of Vo:ztherm: thermodilution cardiac output 4070, hemoglobin concentration 1070, arterial oxygen tension 5070, mixed venous oxygen tension 4070, arterial oxygen saturation 0.3070, and mixed venous oxygen saturation 1070. Coefficient of variation is larger for Vo:ztherm than for Vo:zcart in part because of multiple errors associated with use of multiple component variables in the calculation of Vo:ztherm. Potential sources of error for Vo:zcart and Vo:ztherm include systematic and random error, and are discussed in the manuscript. These sources of error are not mutually exclusive. A further source of error applies to Vo:ztherm but not to Vo:zcart in that Vo:ztherm is mathematically coupled to change in hemoglobin after blood transfusion. We conclude that Vo:zeart is preferred for mea" suring changes in V02 , because Vo:zeart is more accurate and has greater power to detect change than V02therm within the conditions of validation testing for the metabolic cart, and Vo:zcart is not subject to error from mathematical coupling after blood transfusion. We believe that VO:z did not change after blood transfusion as determined by V02cart given that the coefficient of variation for Vo:zcart
979
was low enough to test for a change in VOl of 10070 with a power of 99070. We next examine the two methodologies for DOl' Accuracy of component variables of Dostherm and Do.cart is discussed previously with respect to accuracy of the component variables of Vo1therm and V01cart. Reproducibility of Dostherm has estimated coefficient of variation of 4%, and reproducibility of Do-cart has estimated coefficient of variation of 8%, using principles for calculating the variances of a quotient and product of independent variables as we discussed in the manuscript. As with comparing coefficients of variation for the two methods of determining VOl' coefficient of variation is larger for Do.cart than for Dostherm in part because there is more error associated with use of more component variables in the calculation of Do.cart. Potential sources of error for Dostherm and Do-cart include systematic and random error, and are discussed in the manuscript. Again, these sources of error are not mutually exclusive. Because both Dostherm and Do-cart are calculated using hemoglobin as a component variable, both Dostherm and Doscart are liable to error from mathematical coupling from change in hemoglobin. Thus, neither Dostherm nor Do-cart are free from error due to mathematical coupling. We believe that Dostherm is preferred for measuring changes in DOl since DOl therm has a smaller coefficient of variation than Do-cart. As such, although Do.cart with the larger coefficient of variation did not change, Dostherm with the smaller coefficient of variation increased significantly after blood transfusion with a mean increase of 24%. Furthermore, there was a significant overall increase in mixed venous oxygen saturation after blood transfusion which is physiologically supportive of the conclusion that overall there was an increase in DOl after blood transfusion, and this increase in mixed venous oxygen saturation is not mathematically coupled to change in hemoglobin. We conclude that DOl was increased following blood transfusion. Therefore, V01cart and Do.therm are preferred methods for assessing VOl and DOl, respectively, based on these considerations of accuracy and power to detect a change in the measurement. Moreover, assessment of the relationship between VOl and DOl using V01cart and Do.therm avoids error from mathematical coupling. We conclude that VOl did not change following a significant change in DOl after blood transfusion. We approach the question of whether cardiac output changed after blood transfusion using arguments similar to those comparing Dostherm and Do-cart. Although Fick cardiac output (V01cart divided by arterial-mixed venous oxygen content difference) had
larger coefficient of variation (8%) than thermodilution cardiac output (4%), Fick cardiac output decreased following blood transfusion whereas thermodilution cardiac output did not change. Potential sources of systematic overestimation of thermodilution cardiac output following blood transfusion include loss of indicator signal if tricuspid regurgitation developed after increase in right atrial and pulmonary artery pressures, and changes in thermal conductance of blood or the pulmonary artery catheter following blood transfusion. Potential sources of systematic underestimation of Fick cardiac output include underestimation of Vo1cart and underestimation of mixed venous oxygen saturation after blood transfusion. A further source of error applies to Fick cardiac output but not to thermodilution cardiac output in that Fick cardiac output is mathematically coupled to change in hemoglobin after blood transfusion. We believe that cardiac output did not change after blood transfusion given that the coefficient of variation for thermodilution cardiac output was low enough to test for a change in cardiac output of 10% with a power of 99%. Moreover, thermodilution cardiac output is not mathematically coupled to the increase in hemoglobin following blood transfusion. The decrease in Fick cardiac output is likely artifactual due to mathematical coupling from the significant and substantial increase in the component variable, hemoglobin concentration. Again, mathematical coupling can explain a systematic effect in a relationship when component variables are shared between dependent and independent variables of the relationship. In summary, we conclude that 001 increased following blood transfusion and that VOl did not change after blood transfusion in our patients who had severe ARDS. Finding increase in Vo1therm measured simultaneously to absence of change in Vo1cart leads to our speculation that the finding of pathologic dependence of oxygen consumption on oxygen delivery in previous studies of patients who had ARDS could be due to methodologic error.
J. J. RONCO, M.D. P. T. PRANG, M.D. K. R. WALLEY, M.D. B. R. WIGGS, M.Sc. J. C. FENWICK, M.D. J. A. RUSSELL, M.D. Program of Critical Care Medicine and Pulmonary Research Laboratory St. Paul's Hospital University of British Columbia Vancouver, British Columbia, Canada
ERRATA Please note the following correction to "Studies in the Genetics of Obstructive Sleep Apnea: Familial Aggregation of Symptoms Associated with Sleep-related Breathing Disturbances" (Am Rev Respir Dis 1992; 145:440-444). The author byline should read:
SUSAN REDLINE, TOR TOSTESON, PETER V. TISHLER, MARY A. CARSKADON, and RICHARD P. MILLMAN
tween O, delivery and O, consumption in the adult respiratory distress syndrome. Chest 1983; 81:267-71. 3. Kariman K, Buins 8R. Regulationof tissue oxygenextraction is disturbed in adult respiratory distress syndrome. Am Rev Respir Dis 1985;132:109-14.
To the Editor: Ronco and coworkers, (American Review of Respiratory Disease 1991; 143:1267-1273) concluded in their paper "Oxygen Consumption Is Independent of Changes in Oxygen Delivery in Severe Adult Respiratory Distress Syndrome" that the finding of pathologic dependence of oxygen consumption on oxygen delivery in ARDS in previous studies (1, 2, 3) may be due to methodological error. We have the following fundamental problem with their study: (1) The authors used an argument that cardiac output as measured by thermodilution has a systematic overestimation of the true cardiac output and, therefore, explained the apparent mathematical coupling of calculated VO:z and Oo:z. However, they then argued that the random variation in thermodilution cardiac output and the respective 00:z was less pronounced than those calculated by Fick technique [Le., 00:z = Cao, (Vo:z/Cao:z - Cvo:z)]. Therefore, the thermodilution-derived i>o:z should be assumed to be the more precise values. Patients 1 and 2 in their study had an approximately 24 and 17070 increase in oxygen delivery respectively and demonstrated no change in the calculated VO:z or in measured Vo:z. Therefore, there was no evidence of oxygen supply dependency in these two patients. The question arises that if the mathematical coupling plays a major role in oxygen supply dependency calculation, why didn't these two patients exhibit any increase in the calculated VO:z? These two patients, plus previous studies (1, 3) confirm that there are patients who exhibit no oxygen supply dependency (with either method), despite the significant increases in oxygen delivery. Does mathematical coupling discriminate against particular patients? (2) Ronco and colleagues indicated the excellent precision of the measured VO:z (1.8070), Caos, and Cvo, in their paper. However, if the variation in these parameters is random, why was a systematic reduction of 13070 in the Fick-derived cardiac output observed (from 6.9 l/min to 6.0 IImin) after blood transfusion? Although Ronco and coworkers try to address this issue in their discussion, it is very apparent that they failed to convince at least this reader that the variation of thermodilution in cardiac output will be any less than the Fick cardiac output. Even if we accept that the random variation in measurement of Cao, and Cvo, and measured oxygen consumption were in the same direction, which statistically will be extremely unlikely, one would have difficulty explaining the 13070 reduction in the derived cardiac output and no change in oxygen delivery. Therefore, we do not accept their speculation that the finding of pathologic dependence of oxygen consumption on oxygen delivery in patients with ARDS in previous studies was due to only methodological error. ZAB MOHSENIFAR, M.D. Director Respiratory Therapy Associate Professor of Medicine DAVID J. Ross, M.D. Medical Director Lung Transplantation Assistant Professor of Medicine Cedars-Sinai Medical Center Los Angeles, CA 1. Danek 81, Lynch IP, Weg10, Dantzker DR. The dependence of oxygen uptake on oxygen delivery in the adult respiratory distress syndrome. Am Rev Respir Dis 1980; 122:387-95. 2. Mohsenifar Z, Ooldbach P, Tashkin DP, Campisi Dl, Relationship be978
From the Authors: We interpret the concerns of Drs. Mohsenifar and Ross as, "In evaluating the relationship between oxygen consumption and delivery, how do we justify use of oxygen consumption measured using a metabolic cart (Vo:zcart) while using a different methodology for oxygen delivery based on thermodilution cardiac output (Dostherm)?" We justify our choice of using apparently contrasting methodologies based on considerations of accuracy of the measurement, power to detect a change in the measurement, and by examining potential sources of measurement error for the two methods of determining VO:z, Le.,Vo:zcart compared with Vo:ztherm determined from calculations based on thermodilution cardiac output, and for the two methods of determining Dos, i.e., Dostherm compared with Do-cart determined from calculations based on Fick cardiac output. As we will show, Vo:zcart is preferred over Vo:ztherm because of greater accuracy and power to detect changes in VO:z and is not subject to error from mathematical coupling related to change in hemoglobin after blood transfusion. In contrast, Dostherm is preferred over Do-cart because of considerations of power to detect changes in Do.. Further to preference of Vo:zcart and Dostherm based on considerations of accuracy and power to detect change is that the relationship between Vo:zcart and Dostherm is not compromised by error from mathematical coupling. We believe that we are justified in concluding that VO:z did not change after a significant change in Do, after blood transfusion. We first assess accuracy and power to detect change of the two methodologies for VO:z. We performed validation of Vo:zcart over a range of ventilator settings used for patients who have adult respiratory distress syndrome (ARDS) using an in vitro system whereby we simulated known VO:z using dilution of F'Io:z with precisely regulated flow of N:z (J Crit Care 1991, 6:36). Vo:zcart had an overall measurement error of 1.8070 up to F'Io:z 0.80, minute ventilation 20 L/min, and peak inspiratory pressures of 80 em H:zO. Coefficient of variation of in vitro measurement of Vo:zcart in patients who had ARDS ranged from 4070 at F'Io:z < 0.55 to 7070 at F'Io:z 0.80. We did not perform tests of accuracy of Vo:ztherm. However, accuracy of thermodilution cardiac output is 4 to 10070 as reviewed by others (Am Rev Resp Dis 1982; 126:1001). Using principles for calculating the variances of a product of independent variables (Introduction to the theory of statistics, 3rd edition, Toronto: McGraw-Hill, 1974; 180), we estimate coefficient of variation for Vo:ztherm to be 8070 based on means of our measurements of coefficients of variation for components of Vo:ztherm: thermodilution cardiac output 4070, hemoglobin concentration 1070, arterial oxygen tension 5070, mixed venous oxygen tension 4070, arterial oxygen saturation 0.3070, and mixed venous oxygen saturation 1070. Coefficient of variation is larger for Vo:ztherm than for Vo:zcart in part because of multiple errors associated with use of multiple component variables in the calculation of Vo:ztherm. Potential sources of error for Vo:zcart and Vo:ztherm include systematic and random error, and are discussed in the manuscript. These sources of error are not mutually exclusive. A further source of error applies to Vo:ztherm but not to Vo:zcart in that Vo:ztherm is mathematically coupled to change in hemoglobin after blood transfusion. We conclude that Vo:zeart is preferred for mea" suring changes in V02 , because Vo:zeart is more accurate and has greater power to detect change than V02therm within the conditions of validation testing for the metabolic cart, and Vo:zcart is not subject to error from mathematical coupling after blood transfusion. We believe that VO:z did not change after blood transfusion as determined by V02cart given that the coefficient of variation for Vo:zcart
979
was low enough to test for a change in VOl of 10070 with a power of 99070. We next examine the two methodologies for DOl' Accuracy of component variables of Dostherm and Do.cart is discussed previously with respect to accuracy of the component variables of Vo1therm and V01cart. Reproducibility of Dostherm has estimated coefficient of variation of 4%, and reproducibility of Do-cart has estimated coefficient of variation of 8%, using principles for calculating the variances of a quotient and product of independent variables as we discussed in the manuscript. As with comparing coefficients of variation for the two methods of determining VOl' coefficient of variation is larger for Do.cart than for Dostherm in part because there is more error associated with use of more component variables in the calculation of Do.cart. Potential sources of error for Dostherm and Do-cart include systematic and random error, and are discussed in the manuscript. Again, these sources of error are not mutually exclusive. Because both Dostherm and Do-cart are calculated using hemoglobin as a component variable, both Dostherm and Doscart are liable to error from mathematical coupling from change in hemoglobin. Thus, neither Dostherm nor Do-cart are free from error due to mathematical coupling. We believe that Dostherm is preferred for measuring changes in DOl since DOl therm has a smaller coefficient of variation than Do-cart. As such, although Do.cart with the larger coefficient of variation did not change, Dostherm with the smaller coefficient of variation increased significantly after blood transfusion with a mean increase of 24%. Furthermore, there was a significant overall increase in mixed venous oxygen saturation after blood transfusion which is physiologically supportive of the conclusion that overall there was an increase in DOl after blood transfusion, and this increase in mixed venous oxygen saturation is not mathematically coupled to change in hemoglobin. We conclude that DOl was increased following blood transfusion. Therefore, V01cart and Do.therm are preferred methods for assessing VOl and DOl, respectively, based on these considerations of accuracy and power to detect a change in the measurement. Moreover, assessment of the relationship between VOl and DOl using V01cart and Do.therm avoids error from mathematical coupling. We conclude that VOl did not change following a significant change in DOl after blood transfusion. We approach the question of whether cardiac output changed after blood transfusion using arguments similar to those comparing Dostherm and Do-cart. Although Fick cardiac output (V01cart divided by arterial-mixed venous oxygen content difference) had
larger coefficient of variation (8%) than thermodilution cardiac output (4%), Fick cardiac output decreased following blood transfusion whereas thermodilution cardiac output did not change. Potential sources of systematic overestimation of thermodilution cardiac output following blood transfusion include loss of indicator signal if tricuspid regurgitation developed after increase in right atrial and pulmonary artery pressures, and changes in thermal conductance of blood or the pulmonary artery catheter following blood transfusion. Potential sources of systematic underestimation of Fick cardiac output include underestimation of Vo1cart and underestimation of mixed venous oxygen saturation after blood transfusion. A further source of error applies to Fick cardiac output but not to thermodilution cardiac output in that Fick cardiac output is mathematically coupled to change in hemoglobin after blood transfusion. We believe that cardiac output did not change after blood transfusion given that the coefficient of variation for thermodilution cardiac output was low enough to test for a change in cardiac output of 10% with a power of 99%. Moreover, thermodilution cardiac output is not mathematically coupled to the increase in hemoglobin following blood transfusion. The decrease in Fick cardiac output is likely artifactual due to mathematical coupling from the significant and substantial increase in the component variable, hemoglobin concentration. Again, mathematical coupling can explain a systematic effect in a relationship when component variables are shared between dependent and independent variables of the relationship. In summary, we conclude that 001 increased following blood transfusion and that VOl did not change after blood transfusion in our patients who had severe ARDS. Finding increase in Vo1therm measured simultaneously to absence of change in Vo1cart leads to our speculation that the finding of pathologic dependence of oxygen consumption on oxygen delivery in previous studies of patients who had ARDS could be due to methodologic error.
J. J. RONCO, M.D. P. T. PRANG, M.D. K. R. WALLEY, M.D. B. R. WIGGS, M.Sc. J. C. FENWICK, M.D. J. A. RUSSELL, M.D. Program of Critical Care Medicine and Pulmonary Research Laboratory St. Paul's Hospital University of British Columbia Vancouver, British Columbia, Canada
ERRATA Please note the following correction to "Studies in the Genetics of Obstructive Sleep Apnea: Familial Aggregation of Symptoms Associated with Sleep-related Breathing Disturbances" (Am Rev Respir Dis 1992; 145:440-444). The author byline should read:
SUSAN REDLINE, TOR TOSTESON, PETER V. TISHLER, MARY A. CARSKADON, and RICHARD P. MILLMAN
