ASAIO Journal 2014
Adult Circulatory Support
Association of HeartMate II Left Ventricular Assist Device Flow Estimate with Thermodilution Cardiac Output Tal Hasin,* Marianne Huebner,† Zhuo Li,† Daniel Brown,‡ John M. Stulak,§ Barry A. Boilson,* Lyle Joyce,§ Naveen L. Pereira,* Sudhir S. Kushwaha,* and Soon J. Park§¶
Cardiac output (CO) assessment is important in treating patients with heart failure. Durable left ventricular assist devices (LVADs) provide essentially all CO. In currently used LVADs, estimated device flow is generated by a computerized algorithm. However, LVAD flow estimate may be inaccurate in tracking true CO. We correlated LVAD (HeartMate II) flow with thermodilution CO during postoperative care (day 2–10 after implant) in 81 patients (5,616 paired measurements). Left ventricular assist device flow and CO correlated with a low correlation coefficient (r = 0.42). Left ventricular assist device readings were lower than CO measurements by approximately 0.36 L/min, trending for larger difference with higher values. Left ventricular assist device flow measurements showed less temporal variability compared with CO. Grouping for simultaneous measured blood pressure (BP < 60, 60–70, 70–80, 80–90, and ≥90), the correlation of CO with LVAD flow differed (R = 0.42, 0.67, 0.48, 0.32, 0.32, respectively). Indicating better correlation when mean blood pressure is 60 to 70 mm Hg. Left ventricular assist device flow generally trends with measured CO, but large variability exists, hence flow measures should not be assumed to equal with CO. Clinicians should take into account variables such as high CO, BP, and opening of the aortic valve when interpreting LVAD flow readout. Direct flow sensors incorporated in the LVAD system may allow for better estimation. ASAIO Journal 2014; 60:513–518. Key Words: Heart assist device, cardiac output, flow, blood pressure
revolving rotor, current LVADs empty the left ventricle and pump blood to the ascending aorta. The amount of flow generated by the device is controlled by the speed of rotation and the later can be adjusted by the healthcare provider. Other factors influencing pump flow include the preload, pump work efficiency, and afterload (systemic resistance). Currently used LVADs such as the HeartMate II (Pleasanton, CA) or the HeartWare (Framingham, MA) devices do not have an incorporated flow meter within the device. Estimated flow readout is generated by a computerized algorithm incorporated into the device calculating the power needed to generate the set speed. While patients are routinely followed, the estimated LVAD flow is a readily available parameter to be incorporated among the clinical evaluation of these often clinically challenging patients. However, LVAD flow estimate may be inaccurate in tracking the true cardiac output (CO). Continuous thermodilution CO measurement using the appropriate pulmonary artery catheter (PAC) has been well validated in the postoperative setting and is used as the gold standard for CO assessment.3 In patients implanted with the pulsatile HeartMate XVE device that had an incorporated flow meter, a good association was found with continuous CO measurement, despite a 500 ml bias for the CO.4 During the initial postoperative days after LVAD implant, we routinely use invasive hemodynamic monitoring including continuous CO measurements. Our aim was to correlate LVAD flow estimate with thermodilution CO in patients after LVAD implantation and to isolate clinical circumstances in which this association is weaker or stronger.
Patients with advanced heart failure may benefit from implan-
Methods
tation of left ventricular assist devices (LVADs).1 Newer continuous flow devices offer long-term durability2 and therefore patients may be supported for many years. Using a rapidly
Patients Consecutive patients implanted with a HeartMate II continuous flow device (Thoratec, Pleasanton, CA) either as bridge to transplant (BTT) or destination therapy (DT) from February 2007 to November 2010 in our single center were evaluated retrospectively. Results were analyzed for patients with available simultaneous records of PAC assessed CO and LVAD flow within 2 to 10 days after the surgery. All patients underwent a detailed clinical evaluation before LVAD implant including history, medication review, physical examination, laboratory evaluation, echocardiography, and cardiac catheterization. Outcome measures including length of hospital stay, disposition, and all-cause mortality were recorded. The study was approved by the Institutional Review Board.
From the *Division of Cardiovascular Diseases, †Department of Health Sciences Research, ‡Department of Anesthesiology, and §Division of Cardiothoracic Surgery, Mayo Clinic, Rochester, Minnesota; and ¶Division of Cardiac Surgery University Hospitals Case Medical Center, Cleveland, Ohio. Disclosure: T.H. and S.P. declare accepting a nonrestricted clinical research grant from Thoratec to study endothelial function post-LVAD. S.P. declares consulting for Thoratec. Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML and PDF versions of this article on the journal’s Web site (www.asaiojournal.com). Correspondence: Soon J. Park, MD, Division of Cardiac Surgery University Hospitals Case Medical Center 11100 Euclid Ave, Cleveland, Ohio 44106. Email: [email protected] Copyright © 2014 by the American Society for Artificial Internal Organs
Hemodynamic Measurements Postoperative CO measurement was performed using VIP PAC (7.5F) and the Vigilance II monitor (Edwards Lifesciences,
DOI: 10.1097/MAT.0000000000000119
513
514 HASIN et al. Irvine, CA). Left ventricular assist device flow and CO measurements were recorded among other vital signs by the attending nurse every hour for as long as the patient was not ambulating and thereafter every 4 hour. All recorded values were accessible and were electronically derived. Output or flow values 8 (L/min) were considered unreliable and were removed from the analysis. Simultaneously measured arterial line derived blood pressure was also recorded and used in our analysis. Statistical Analysis Categorical variables were summarized as frequency and percent while continuous variables were summarized as mean (standard deviation) or median (quartiles) as appropriate. Left ventricular assist device flow and CO were reported as mean and standard deviation. The values of these two variables at different time points were presented by scatter plots, for the whole group, by hypertension subgroups, and for each individual patient, respectively. Deming regression analysis allows for the presence of measurement errors in both LVAD flow and CO and was used to calculate a fitted line. Correlations between the CO and LVAD flow measurements were evaluated using Pearson correlation coefficients. However, although correlation coefficients measure the strength of the relationships between the two cardiac measures, they do not measure agreement. Bland–Altman plots were used to measure agreement between LVAD flow and CO. Bland– Altman plots are created by plotting the mean of the two measurements against the difference for each subject at each time point. Lines were added to represent the bias (solid line) between the two measurements across all subjects and the limits of agreement (dashed lines). These lines were calculated taking into account repeated measures for each patient. It is expected that 95% of the differences lie within the lower and upper limits of agreement. A level of 0.05 was considered statistically significant. All analyses were performed using R version 2.15.05 and SAS v.9.2 (SAS Institute, Inc., Cary, NC). Results Eighty-one patients were included in the study. Patient characteristics are shown in Table 1. Mean age was 62 years, 85% were males, 53% had ischemic etiology, and 33% were implanted as BTT. Thirty patients (37%) had a history of hypertension. Corrective surgery for tricuspid regurgitation was performed concomitant to LVAD implant in 39 patients and aortic valve procedure (usually aortic valve approximation stitch) in 7. The patients required invasive monitoring using a PAC for 5 (3;7) (median, [IQR]) days. At least some opening of the aortic valve was noted in 31% of the patients during the first postoperative echo performed 6 (4;10) (median, [IQR]) days after implant. None of the patients had significant (more than mild) aortic incompetence. Patients were discharged after (median) 18 days, 63% were discharged to a rehabilitation facility and 7% died during hospitalization. During 1.6 (0.7;2.4) (median, [IQR]) years follow-up, 18 patients were transplanted and 25 died while on LVAD support. One year survival was 85% (95% CI, 78–94) and 3 year survival was 61% (95% CI, 48–77).
Table 1. Patient Characteristics Variable, n = 81 Demographic Age (years) (mean, SD) Gender (male) Hypertension Diabetes mellitus Ischemic etiology Bridge to transplant Preoperative Heart rate (beats/min) Weight (kg) Blood pressure, systolic (mm Hg) Blood pressure, diastolic (mm Hg) NYHA class Angiotensin blocker Beta receptor blockers Hemoglobin (g%) NT-proBNP (pg/mL), n = 50 Creatinine (mg%) INTERMACS score Echocardiography preoperative Left ventricular diastolic diameter (mm) Ejection fraction (%) Catheterization preoperative Mean right atrial pressure (mm Hg), n = 76 Mean pulmonary pressure (mm Hg), n = 76 Mean wedge pressure (mm Hg), n = 73 RV stroke work index (g/m), n = 76 Cardiac index (L/min/m2), n = 76 Operation and perioperative Bypass time (min) Tricuspid valve procedure Aortic valve procedure Duration of inotropic support (hr) Last PAC measurement (days postoperative) LVAD RPM at last PAC measurement, n = 79 EF (%) in first postoperative echo, n = 74 Aortic incompetence, first echo, n = 78 AV opening (any), first echo, n = 77 Tricuspid regurgitation >moderate, first echo, n = 76 Duration of hospitalization (days)
Mean ± SD/Median (25th%;75th%) 62 ± 11 69 (85%) 30 (37%) 27 (33%) 43 (53%) 27 (33%) 77 ± 17 88 ± 18 101 ± 12 64 ± 8 III: 27 (33%) IV: 54 (67%) 51 (63%) 69 (85%) 11.9 ± 1.8 4,510 (2,249;8,353) 1.6 ± 0.7 3.9 ± 1.6 67 ± 8 20 ± 8 15.8 ± 6.6 40.0 ± 8.4 23.7 ± 6.6 7.5 ± 3.8 1.9 ± 0.5 110 ± 43 39 (48%) 7 (9%) 120 (72;234) 5 (3;7) 9,400 (9,200;9,400) 23 (17;30) None 30 (39%) Trivial/mild (61%) 24 (31%) 11 (14%) 18 (12;28)
Baseline and early postoperative characteristics of 81 patients implanted with a HeartMate II LVAD for whom cardiac output was compared with LVAD flow. LVAD, left ventricular assist device; NYHA, New York Heart Association class; Angiotensin blocker, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker; RV, right ventricle, RV stroke work index equals (0.0136 × (MPAP − RAP) × stroke volume index); PAC, pulmonary artery catheter; AV, aortic valve; INTERMACS score as previously published.
We analyzed 5,616 paired LVAD flow–CO measurements of these we excluded 419 in which values were below 2 or above 8 L/min leaving 5,197 paired measurements for analysis, median 54 measurements/patients (range 1–206). The distributions of CO and LVAD flow values are shown in Figure 1. Left ventricular assist device flow values were 5.1 ± 0.7 (mean ± SD, L/min) and PAC CO values 5.5 ± 1.0 (mean ± SD, L/min). The correlation between LVAD flow and PAC CO is shown in Figure 2. Left ventricular assist device and CO measurements were overall correlated but with a low correlation coefficient
515
LVAD FLOW ESTIMATE EVALUATION
Figure 1. Distributions of pulmonary artery catheter measured cardiac output and left ventricular assist device (LVAD) flow values measured during the study period for all LVAD supported patients .
(r = 0.40). Overall LVAD readings tended to be lower than the corresponding PAC CO measurements. The Bland–Altman plot is shown in Figure 3. The region of agreement for the difference CO–LVAD flow was −1.87 to 2.59 with mean 0.36 (SD = 0.4). As 96.5% of the data points were within the limits of agreement (2% were below and 1.5% were above), it appears that overall there is agreement between the two measurements, despite a large variability. A trend was noted for more increase in CO–LVAD flow for higher mean values. A subgroup analysis was performed for LVAD flow outside the mid-range of 4 to 6 L/min. Analyzing for 149 paired measurements from 18 patients with flow between 3 and 4 L/min correlation coefficient is 0.14, p = 0.084. In 482 paired measurements from 40 patients with flow between 6 and 8 L/min correlation coefficient is −0.11, p = 0.012 (see Supplemental Digital Content, http://links.lww.com/ASAIO/A51).
Comparison of measurements over time showed less variation in the LVAD flow measurements compared with the CO. The ratio of within subject standard deviation of CO to LVAD flow was constant (mean 1.7) indicating larger variability for CO. Selected examples of the variability in measurements are shown in Figure 4. Simultaneous blood pressure, CO, and LVAD flow measurements were available for 4,898 data points in 81 patients. Fortyeight data points (in 16 patients) with mean BP
Adult Circulatory Support
Association of HeartMate II Left Ventricular Assist Device Flow Estimate with Thermodilution Cardiac Output Tal Hasin,* Marianne Huebner,† Zhuo Li,† Daniel Brown,‡ John M. Stulak,§ Barry A. Boilson,* Lyle Joyce,§ Naveen L. Pereira,* Sudhir S. Kushwaha,* and Soon J. Park§¶
Cardiac output (CO) assessment is important in treating patients with heart failure. Durable left ventricular assist devices (LVADs) provide essentially all CO. In currently used LVADs, estimated device flow is generated by a computerized algorithm. However, LVAD flow estimate may be inaccurate in tracking true CO. We correlated LVAD (HeartMate II) flow with thermodilution CO during postoperative care (day 2–10 after implant) in 81 patients (5,616 paired measurements). Left ventricular assist device flow and CO correlated with a low correlation coefficient (r = 0.42). Left ventricular assist device readings were lower than CO measurements by approximately 0.36 L/min, trending for larger difference with higher values. Left ventricular assist device flow measurements showed less temporal variability compared with CO. Grouping for simultaneous measured blood pressure (BP < 60, 60–70, 70–80, 80–90, and ≥90), the correlation of CO with LVAD flow differed (R = 0.42, 0.67, 0.48, 0.32, 0.32, respectively). Indicating better correlation when mean blood pressure is 60 to 70 mm Hg. Left ventricular assist device flow generally trends with measured CO, but large variability exists, hence flow measures should not be assumed to equal with CO. Clinicians should take into account variables such as high CO, BP, and opening of the aortic valve when interpreting LVAD flow readout. Direct flow sensors incorporated in the LVAD system may allow for better estimation. ASAIO Journal 2014; 60:513–518. Key Words: Heart assist device, cardiac output, flow, blood pressure
revolving rotor, current LVADs empty the left ventricle and pump blood to the ascending aorta. The amount of flow generated by the device is controlled by the speed of rotation and the later can be adjusted by the healthcare provider. Other factors influencing pump flow include the preload, pump work efficiency, and afterload (systemic resistance). Currently used LVADs such as the HeartMate II (Pleasanton, CA) or the HeartWare (Framingham, MA) devices do not have an incorporated flow meter within the device. Estimated flow readout is generated by a computerized algorithm incorporated into the device calculating the power needed to generate the set speed. While patients are routinely followed, the estimated LVAD flow is a readily available parameter to be incorporated among the clinical evaluation of these often clinically challenging patients. However, LVAD flow estimate may be inaccurate in tracking the true cardiac output (CO). Continuous thermodilution CO measurement using the appropriate pulmonary artery catheter (PAC) has been well validated in the postoperative setting and is used as the gold standard for CO assessment.3 In patients implanted with the pulsatile HeartMate XVE device that had an incorporated flow meter, a good association was found with continuous CO measurement, despite a 500 ml bias for the CO.4 During the initial postoperative days after LVAD implant, we routinely use invasive hemodynamic monitoring including continuous CO measurements. Our aim was to correlate LVAD flow estimate with thermodilution CO in patients after LVAD implantation and to isolate clinical circumstances in which this association is weaker or stronger.
Patients with advanced heart failure may benefit from implan-
Methods
tation of left ventricular assist devices (LVADs).1 Newer continuous flow devices offer long-term durability2 and therefore patients may be supported for many years. Using a rapidly
Patients Consecutive patients implanted with a HeartMate II continuous flow device (Thoratec, Pleasanton, CA) either as bridge to transplant (BTT) or destination therapy (DT) from February 2007 to November 2010 in our single center were evaluated retrospectively. Results were analyzed for patients with available simultaneous records of PAC assessed CO and LVAD flow within 2 to 10 days after the surgery. All patients underwent a detailed clinical evaluation before LVAD implant including history, medication review, physical examination, laboratory evaluation, echocardiography, and cardiac catheterization. Outcome measures including length of hospital stay, disposition, and all-cause mortality were recorded. The study was approved by the Institutional Review Board.
From the *Division of Cardiovascular Diseases, †Department of Health Sciences Research, ‡Department of Anesthesiology, and §Division of Cardiothoracic Surgery, Mayo Clinic, Rochester, Minnesota; and ¶Division of Cardiac Surgery University Hospitals Case Medical Center, Cleveland, Ohio. Disclosure: T.H. and S.P. declare accepting a nonrestricted clinical research grant from Thoratec to study endothelial function post-LVAD. S.P. declares consulting for Thoratec. Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML and PDF versions of this article on the journal’s Web site (www.asaiojournal.com). Correspondence: Soon J. Park, MD, Division of Cardiac Surgery University Hospitals Case Medical Center 11100 Euclid Ave, Cleveland, Ohio 44106. Email: [email protected] Copyright © 2014 by the American Society for Artificial Internal Organs
Hemodynamic Measurements Postoperative CO measurement was performed using VIP PAC (7.5F) and the Vigilance II monitor (Edwards Lifesciences,
DOI: 10.1097/MAT.0000000000000119
513
514 HASIN et al. Irvine, CA). Left ventricular assist device flow and CO measurements were recorded among other vital signs by the attending nurse every hour for as long as the patient was not ambulating and thereafter every 4 hour. All recorded values were accessible and were electronically derived. Output or flow values 8 (L/min) were considered unreliable and were removed from the analysis. Simultaneously measured arterial line derived blood pressure was also recorded and used in our analysis. Statistical Analysis Categorical variables were summarized as frequency and percent while continuous variables were summarized as mean (standard deviation) or median (quartiles) as appropriate. Left ventricular assist device flow and CO were reported as mean and standard deviation. The values of these two variables at different time points were presented by scatter plots, for the whole group, by hypertension subgroups, and for each individual patient, respectively. Deming regression analysis allows for the presence of measurement errors in both LVAD flow and CO and was used to calculate a fitted line. Correlations between the CO and LVAD flow measurements were evaluated using Pearson correlation coefficients. However, although correlation coefficients measure the strength of the relationships between the two cardiac measures, they do not measure agreement. Bland–Altman plots were used to measure agreement between LVAD flow and CO. Bland– Altman plots are created by plotting the mean of the two measurements against the difference for each subject at each time point. Lines were added to represent the bias (solid line) between the two measurements across all subjects and the limits of agreement (dashed lines). These lines were calculated taking into account repeated measures for each patient. It is expected that 95% of the differences lie within the lower and upper limits of agreement. A level of 0.05 was considered statistically significant. All analyses were performed using R version 2.15.05 and SAS v.9.2 (SAS Institute, Inc., Cary, NC). Results Eighty-one patients were included in the study. Patient characteristics are shown in Table 1. Mean age was 62 years, 85% were males, 53% had ischemic etiology, and 33% were implanted as BTT. Thirty patients (37%) had a history of hypertension. Corrective surgery for tricuspid regurgitation was performed concomitant to LVAD implant in 39 patients and aortic valve procedure (usually aortic valve approximation stitch) in 7. The patients required invasive monitoring using a PAC for 5 (3;7) (median, [IQR]) days. At least some opening of the aortic valve was noted in 31% of the patients during the first postoperative echo performed 6 (4;10) (median, [IQR]) days after implant. None of the patients had significant (more than mild) aortic incompetence. Patients were discharged after (median) 18 days, 63% were discharged to a rehabilitation facility and 7% died during hospitalization. During 1.6 (0.7;2.4) (median, [IQR]) years follow-up, 18 patients were transplanted and 25 died while on LVAD support. One year survival was 85% (95% CI, 78–94) and 3 year survival was 61% (95% CI, 48–77).
Table 1. Patient Characteristics Variable, n = 81 Demographic Age (years) (mean, SD) Gender (male) Hypertension Diabetes mellitus Ischemic etiology Bridge to transplant Preoperative Heart rate (beats/min) Weight (kg) Blood pressure, systolic (mm Hg) Blood pressure, diastolic (mm Hg) NYHA class Angiotensin blocker Beta receptor blockers Hemoglobin (g%) NT-proBNP (pg/mL), n = 50 Creatinine (mg%) INTERMACS score Echocardiography preoperative Left ventricular diastolic diameter (mm) Ejection fraction (%) Catheterization preoperative Mean right atrial pressure (mm Hg), n = 76 Mean pulmonary pressure (mm Hg), n = 76 Mean wedge pressure (mm Hg), n = 73 RV stroke work index (g/m), n = 76 Cardiac index (L/min/m2), n = 76 Operation and perioperative Bypass time (min) Tricuspid valve procedure Aortic valve procedure Duration of inotropic support (hr) Last PAC measurement (days postoperative) LVAD RPM at last PAC measurement, n = 79 EF (%) in first postoperative echo, n = 74 Aortic incompetence, first echo, n = 78 AV opening (any), first echo, n = 77 Tricuspid regurgitation >moderate, first echo, n = 76 Duration of hospitalization (days)
Mean ± SD/Median (25th%;75th%) 62 ± 11 69 (85%) 30 (37%) 27 (33%) 43 (53%) 27 (33%) 77 ± 17 88 ± 18 101 ± 12 64 ± 8 III: 27 (33%) IV: 54 (67%) 51 (63%) 69 (85%) 11.9 ± 1.8 4,510 (2,249;8,353) 1.6 ± 0.7 3.9 ± 1.6 67 ± 8 20 ± 8 15.8 ± 6.6 40.0 ± 8.4 23.7 ± 6.6 7.5 ± 3.8 1.9 ± 0.5 110 ± 43 39 (48%) 7 (9%) 120 (72;234) 5 (3;7) 9,400 (9,200;9,400) 23 (17;30) None 30 (39%) Trivial/mild (61%) 24 (31%) 11 (14%) 18 (12;28)
Baseline and early postoperative characteristics of 81 patients implanted with a HeartMate II LVAD for whom cardiac output was compared with LVAD flow. LVAD, left ventricular assist device; NYHA, New York Heart Association class; Angiotensin blocker, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker; RV, right ventricle, RV stroke work index equals (0.0136 × (MPAP − RAP) × stroke volume index); PAC, pulmonary artery catheter; AV, aortic valve; INTERMACS score as previously published.
We analyzed 5,616 paired LVAD flow–CO measurements of these we excluded 419 in which values were below 2 or above 8 L/min leaving 5,197 paired measurements for analysis, median 54 measurements/patients (range 1–206). The distributions of CO and LVAD flow values are shown in Figure 1. Left ventricular assist device flow values were 5.1 ± 0.7 (mean ± SD, L/min) and PAC CO values 5.5 ± 1.0 (mean ± SD, L/min). The correlation between LVAD flow and PAC CO is shown in Figure 2. Left ventricular assist device and CO measurements were overall correlated but with a low correlation coefficient
515
LVAD FLOW ESTIMATE EVALUATION
Figure 1. Distributions of pulmonary artery catheter measured cardiac output and left ventricular assist device (LVAD) flow values measured during the study period for all LVAD supported patients .
(r = 0.40). Overall LVAD readings tended to be lower than the corresponding PAC CO measurements. The Bland–Altman plot is shown in Figure 3. The region of agreement for the difference CO–LVAD flow was −1.87 to 2.59 with mean 0.36 (SD = 0.4). As 96.5% of the data points were within the limits of agreement (2% were below and 1.5% were above), it appears that overall there is agreement between the two measurements, despite a large variability. A trend was noted for more increase in CO–LVAD flow for higher mean values. A subgroup analysis was performed for LVAD flow outside the mid-range of 4 to 6 L/min. Analyzing for 149 paired measurements from 18 patients with flow between 3 and 4 L/min correlation coefficient is 0.14, p = 0.084. In 482 paired measurements from 40 patients with flow between 6 and 8 L/min correlation coefficient is −0.11, p = 0.012 (see Supplemental Digital Content, http://links.lww.com/ASAIO/A51).
Comparison of measurements over time showed less variation in the LVAD flow measurements compared with the CO. The ratio of within subject standard deviation of CO to LVAD flow was constant (mean 1.7) indicating larger variability for CO. Selected examples of the variability in measurements are shown in Figure 4. Simultaneous blood pressure, CO, and LVAD flow measurements were available for 4,898 data points in 81 patients. Fortyeight data points (in 16 patients) with mean BP
