bs_bs_banner
Copyright © 2015 International Center for Artificial Organs and Transplantation and Wiley Periodicals, Inc.
Multisite Near Infrared Spectroscopy During Cardiopulmonary Bypass in Pediatric Patients *Zaccaria Ricci, *Roberta Haiberger, †Lorenzo Tofani, ‡Stefano Romagnoli, *Isabella Favia, and *Paola Cogo *Department of Cardiology and Cardiac Surgery, Pediatric Cardiac Intensive Care Unit, Bambino Gesù Children’s Hospital, IRCCS, Rome; †Department of Neurosciences, Psychology, Drug Research and Child Health, University of Florence, ‡Department of Human Health Sciences, Section of Anaesthesiology and Intensive Care, University of Florence, Azienda Ospedaliero-Universitaria Careggi, Florence, Italy
Abstract: Multisite near infrared spectroscopy (NIRS) monitoring during pediatric cardiopulmonary bypass (CPB) has not been extensively validated. Although it might be rational to explore regional tissue saturation at different body sites (namely brain, kidney, upper body, lower body), conflicting results are currently provided by experience in children. The aim of our study was to evaluate absolute values of multisite NIRS saturation during CPB in a cohort of infants undergoing pediatric cardiac surgery to describe average differences between cerebral, renal, upper body (arm), and lower body (thigh) regional saturation. Furthermore, the correlation between cerebral NIRS and cardiac index (CI) at CPB weaning was evaluated. Twenty-five infants were enrolled: their median weight, age, and body surface area were 3.9 (3.3–6) kg, 111 (47–203) days, and 0.24 (0.22–0.33) m2, respectively. Median Aristotle score was 8 (6–10), and vasoactive inotropic score at CPB weaning was 16 (14–25). A total of 17 430 data points were recorded by each sensor: twoway ANOVA showed that time (P < 0.0001) and site (P = 0.0001) significantly affected variations of NIRS values: however, if cerebral NIRS values are excluded, sensor site is no more significant (P = 0.184 in the no cir-
culatory arrest [noCA] group and P = 0.42 in the circulatory arrest [CA] group). Analysis of NIRS saturation changes over time showed that, at all sites, average NIRS values increased after CPB start, even if the increase of cerebral saturation was less intense than other sites (P < 0.0001). Detailed analysis of interaction between site of NIRS measurement and time point showed that cerebral NIRS (ranging from 65 to 75%) was always significantly lower than that of other channels (P < 0.0001) that tended to be in the range of oversaturation (80–90%), especially during the CPB phase. Average cerebral NIRS values of patients who did not undergo circulatory arrest (CA) during CPB, 10 min after CPB weaning, were associated with average CI values with a significant correlation (r = 0.7, P = 0.003). In conclusion, during CPB, cerebral NIRS values are expected to remain constantly lower than somatic sensors, which instead tend to show similar elevated saturations, regardless of their position. Based on these results, positioning of noncerebral NIRS sensors during CPB without CA may be questioned. Key Words: Near-infrared spectroscopy—Pediatric cardiac surgery—Cardiopulmonary bypass—Deep hypothermic circulatory arrest.
Clinical utilization of near infrared spectroscopy (NIRS) has recently increased (1), several years after its first conception (2). In particular, the application of NIRS monitoring in pediatric cardiac surgery has
gained ongoing success due to its reliability as a perfusion assessment tool, ease of use, and low invasiveness (3). NIRS uses principles of light transmission and absorption to noninvasively measure the concentrations of oxygenated (HbO2) and deoxygenated (Hb) hemoglobin in tissue (4). NIRS instruments may be divided into two distinct types: concentration monitors and saturation monitors (4). In vivo saturation spectrometers are those most commonly used and described in clinical practice: they do not attempt to measure concentrations of HbO2 and Hb in tissue, but instead measure the ratio of the two
doi:10.1111/aor.12424 Received May 2014; revised September 2014. Address correspondence and reprint requests to Dr. Zaccaria Ricci, Department of Cardiology, Cardiac Surgery, Pediatric Cardiac Intensive Care Unit, Bambino Gesù Children’s Hospital, IRCCS, Piazza S. Onofrio 4, 00165 Rome, Italy. E-mail: [email protected] Artificial Organs 2015, 39(7):584–590
NIRS DURING PEDIATRIC CARDIAC SURGERY and derive the hemoglobin oxygen saturation (4). Some of these instruments, furthermore, attempt also to reduce the effects of surface tissues overlying the brain by measuring the absorption ratios at two different source-detector spacing and subtracting one from the others (5). Several targets have been proposed as objects of NIRS scanning: cerebral saturation, as a marker of brain perfusion or as a surrogate of superior vena cava oxygen saturation (ScVO2); splanchnic saturation, as a marker of abdominal perfusion; and renal saturation, as an index of renal blood flow adequacy (6). As a matter of fact, currently, NIRS monitoring reliability and application during pediatric cardiac surgery, both in the intraoperative and postoperative phases, are debated (7). In particular, the role of multisite sensor positioning and monitoring has not extensively been evaluated, and its application is limited by the cost of each disposable sensor pad. Cardiopulmonary bypass (CPB) is one of the most delicate phases of open-heart pediatric surgery, due to the risk of cerebral and systemic malperfusion. Although several technical aspects recently developed for pediatric CPB have significantly improved the outcome of children with congenital heart disease after cardiac surgery (8), multiorgan protection during the nonpulsatile phase of CPB is still a primary target. There is still great uncertainty about the management of most important clinical parameters, such as optimal temperature, hematocrit, blood flow, perfusion pressure, metabolic strategy, etc. (8–10). Multisite monitoring of perfusion during CPB could contribute to address this important issue. The aim of our study was to evaluate absolute values of multisite NIRS monitoring during CPB in a cohort of infants undergoing pediatric cardiac surgery to describe average differences between cerebral, renal, upper body (arm), and lower body (thigh) regional saturation. Furthermore, the correlation between cerebral NIRS and cardiac index (CI) at CPB weaning was evaluated. PATIENTS AND METHODS Design A retrospective analysis of prospectively collected data was conducted on infants undergoing pediatric cardiac surgery with CPB with a four-channel NIRS intraoperative recording. The INVOS 5100C Cerebral Oximeter (Somanetics, Troy, MI, USA) was used in all patients. From June to December 2013, to evaluate institutional standardization of intraoperative monitoring, all infants were provided with multisite NIRS after anesthesia induction: four NIRS
585
probes were placed on the right forehead, on the right posterior flank, on the right upper arm, and on the right thigh to estimate the cerebral, renal, and upper and lower body muscle perfusion, respectively. Neonatal probes were used in patients weighing less than 5 kg, and pediatric probes were applied in all other patients. Recorded data were retrieved from the INVOS monitor and utilized for the analysis. The INVOS monitor automatically recorded two data points per minute in all monitored patients. All patients with a number of missing data points exceeding 10% of the overall monitored time were not included in the analysis. Inclusion and exclusion criteria and data collection Inclusion criteria were infants with elective surgery, in stable hemodynamic condition, and younger than 1 year old. Demographic, surgical, and clinical data (Aristotle score, vasoactive inotropic score [VIS], Pediatric Index of Mortality 2 [PIM2], length of mechanical ventilation [MV], pediatric cardiac intensive care unit length of stay [PCICU LOS]) were obtained from an institutional database, and CPB data were retrieved from perfusionists’ electronic database. General anesthetic and CPB management All patients received general anesthesia with sevoflurane at induction and midazolam (0.05 mg/kg/ h), fentanyl (5 μg/kg/h), and rocuronium (0.5 mg/ kg/h) infusion for maintenance. CPB was always primed with packed red blood cells (PRBCs) and 5% albumin, and conducted with alpha stat strategy and blood cardioplegia. A CPB flow of 150 mL/kg/min, a hematocrit of 28–30%, and moderate hypothermia with the temperature ranging between 25 and 32°C were targeted. Deep hypothermic circulatory arrest (DHCA) at 18°C was used when necessary. The standard institutional protocol for CPB weaning consisted of the use of milrinone (0.75 μg/kg/min and dopamine 5 μg/kg/min) started at the end of crossclamp as first choice. If MAP was below 45 mm Hg and filling pressures did not indicate the need for fluid replacement, dopamine was increased to 10 μg/ kg/min. If such targets were not achieved, adrenaline (0.05–0.3 μg/kg/min) was added. Evaluated scores and hemodynamic measurements Aristotle score is currently used in our department to classify surgical risk of our patients (11). VIS was utilized to assess patients’ hemodynamic status at CPB weaning and it was calculated as follows: dopamine dose (μg/kg/min) + dobutamine dose (μg/kg/min) + 100 × epinephrine dose (μg/kg/min) + Artif Organs, Vol. 39, No. 7, 2015
586
Z. RICCI ET AL.
10 × milrinone dose (μg/kg/ min) + 10 000 × Vasopressin dose (units/kg/min) + 100 × norepinephrine dose (μg/kg/min) (12). PIM2 score is utilized in our hospital to assess critical illness severity of children admitted to our pediatric intensive care units (PICUs) (13). Finally, after CPB, CI is routinely monitored at our institution by pressure recording analytical method (PRAM) (14): it is a minimally invasive hemodynamic monitoring system based on mathematical analysis of the invasive arterial waveform recorded at a high sampling rate (1000 Hz). The area under the pressure wave is examined during the whole cardiac cycle, including the postdicrotic notch phase, and it is utilized to assess the patients’ stroke index. PRAM monitoring requires an arterial waveform without artifacts and the correct identification of the dicrotic notch (it cannot be used during CPB and in case of under or overdamping); PRAM monitor automatically records each minute’s CI and makes it available for download. The institutional review board of the Bambino Gesù Children’s Hospital approved the protocol and waived the need for informed consent due to the retrospective nature of the study [Prot n. 235LB]. Study objectives The primary aim of this study was to describe the relative average value of the four regional saturations as measured by NIRS sensors during the main intraoperative phases: from anesthesia induction (after intubation) to sternotomy (baseline), from sternotomy to CPB start (T1), from CPB start to CPB weaning (T2), and from CPB weaning to sternal closure (T3). We also attempted to evaluate if significant differences were present between values recorded by each sensor, during each intraoperative period. Third, we compared the NIRS values in the subgroup of patients undergoing CPB without DHCA (noCA) with those requiring DHCA (CA). In patients undergoing DHCA, this was assessed as an adjunctive intraoperative phase (T2CA) interposed between the phase before (T2pre) and after the arrest (T2post). Finally, we analyzed the reliability of cerebral NIRS as indicator of global perfusion, as surrogate of ScVO2, during CPB weaning by correlating it with CI values: the average cerebral NIRS value and the average CI value of each patient during the first 10 min after CPB weaning were considered. Statistical analysis The Mann–Whitney test or Student’s t-test were used, according to the variable distribution, to compare continuous variables. Two-way analysis of variance (ANOVA) for repeated measures was used Artif Organs, Vol. 39, No. 7, 2015
to assess NIRS changes over the four predefined time points. To assess the association between regional saturation (continuous variable) and other covariates (intraoperative phases [T1, T2, T3, and TCA] and measurement sites), a generalized estimating equation (GEE) multivariable linear regression model was conceived, considering the nature of the data. This choice was made for the ability of GEE to model the dependence structure of the data. In this analysis, an unstructured working correlation has been used. The model also included, in addition to main effects, interactions between the channel and measurement sites. The normality of the residuals was tested to assess the adequacy of the model. Spearman’s correlation was used to verify univariate association between continuous variables. Data are presented as median and interquartile range (IQR) or average and standard deviation (SD) or standard error (SE) where necessary. A P value 0.05, except where indicated by the symbols. * P = 0.028; ** P = 0.044. † Preoperatively. X-Clamp, cross-clamp.
represents the tissue oxygen average saturations at the four sites in the noCA and CA groups: two-way ANOVA showed that time (P < 0.0001 in both groups) and site (P = 0.0001 in the noCA group and P = 0.012 in the CA group) significantly affected variations of NIRS values. Once cerebral NIRS values were excluded from the two-way ANOVA, the impact of sensor site was no longer significant (P = 0.184 in the noCA group and P = 0.42 in the CA group). For this reason, cerebral NIRS was used as reference for channel variable in the GEE linear regression model: it was associated with a significantly different saturation level, at intercept, than that of other channels (always P < 0.0001), with similar estimates between arm (+10.36 [0.6]), flank (+5.88 [1.13]), or thigh (+10.69 [1.05]) saturations (Table 2). Detailed analysis of interaction between site of NIRS measurement and time point showed that cerebral NIRS was always significantly lower than that of other channels (P < 0.0001). Furthermore, at T2 NIRS saturation increased at all sites: cerebral NIRS saturation rise according to GEE was significantly lower with respect to somatic values (P < 0.0001). Finally, during the phase of CA, all saturations were lowest (although still significantly different between cerebral and others) (Table 2).
Hemodynamic monitoring after CPB weaning When average cerebral NIRS values of noCA patients after CPB weaning (T3) were associated
FIG. 1. Average modifications of NIRS over time. Panel A: patients without circulatory arrest (noCA). Panel B: patients undergoing CA. Data are depicted as mean and standard deviation. In panel B, T2CA represents the CA phase and it is interposed between the CPB phases before (T2pre) and after the arrest (T2post). Artif Organs, Vol. 39, No. 7, 2015
588
Z. RICCI ET AL. TABLE 2. GEE of NIRS values at different sites (CH) and time points (T) Analysis of GEE parameter estimates Empirical standard error estimates Parameter Intercept T1 × CH T1 × CH T1 × CH T1 × CH T2 × CH T2 × CH T2 × CH T2 × CH T3 × CH T3 × CH T3 × CH T3 × CH T2CA × CH T2CA × CH T2CA × CH T2CA × CH
2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1
Estimate
SE
83.97 2.55 5.02 2.13 0.00 6.75 13.63 4.67 0.00 9.01 10.10 2.29 0.00 −6.44 −6.25 −6.02 0.00
4.23 0.90 1.06 0.88 0.00 0.64 1.20 1.10 0.00 0.67 1.31 1.15 0.00 1.18 1.36 1.34 0.00
95% conf limits 75.67 0.77 2.93 0.40 0.00 5.49 11.28 2.51 0.00 7.69 7.52 0.03 0.00 −8.77 −8.93 −8.66 0.00
92.27 4.33 7.12 3.86 0.00 8.01 15.99 6.84 0.00 10.33 12.68 4.55 0.00 −4.11 −3.57 −3.39 0.00
Z
P
19.82 2.81 4.70 2.42 — 10.48 11.35 4.24 — 13.36 7.68 1.99 — −5.42 −4.57 −4.49 —
Copyright © 2015 International Center for Artificial Organs and Transplantation and Wiley Periodicals, Inc.
Multisite Near Infrared Spectroscopy During Cardiopulmonary Bypass in Pediatric Patients *Zaccaria Ricci, *Roberta Haiberger, †Lorenzo Tofani, ‡Stefano Romagnoli, *Isabella Favia, and *Paola Cogo *Department of Cardiology and Cardiac Surgery, Pediatric Cardiac Intensive Care Unit, Bambino Gesù Children’s Hospital, IRCCS, Rome; †Department of Neurosciences, Psychology, Drug Research and Child Health, University of Florence, ‡Department of Human Health Sciences, Section of Anaesthesiology and Intensive Care, University of Florence, Azienda Ospedaliero-Universitaria Careggi, Florence, Italy
Abstract: Multisite near infrared spectroscopy (NIRS) monitoring during pediatric cardiopulmonary bypass (CPB) has not been extensively validated. Although it might be rational to explore regional tissue saturation at different body sites (namely brain, kidney, upper body, lower body), conflicting results are currently provided by experience in children. The aim of our study was to evaluate absolute values of multisite NIRS saturation during CPB in a cohort of infants undergoing pediatric cardiac surgery to describe average differences between cerebral, renal, upper body (arm), and lower body (thigh) regional saturation. Furthermore, the correlation between cerebral NIRS and cardiac index (CI) at CPB weaning was evaluated. Twenty-five infants were enrolled: their median weight, age, and body surface area were 3.9 (3.3–6) kg, 111 (47–203) days, and 0.24 (0.22–0.33) m2, respectively. Median Aristotle score was 8 (6–10), and vasoactive inotropic score at CPB weaning was 16 (14–25). A total of 17 430 data points were recorded by each sensor: twoway ANOVA showed that time (P < 0.0001) and site (P = 0.0001) significantly affected variations of NIRS values: however, if cerebral NIRS values are excluded, sensor site is no more significant (P = 0.184 in the no cir-
culatory arrest [noCA] group and P = 0.42 in the circulatory arrest [CA] group). Analysis of NIRS saturation changes over time showed that, at all sites, average NIRS values increased after CPB start, even if the increase of cerebral saturation was less intense than other sites (P < 0.0001). Detailed analysis of interaction between site of NIRS measurement and time point showed that cerebral NIRS (ranging from 65 to 75%) was always significantly lower than that of other channels (P < 0.0001) that tended to be in the range of oversaturation (80–90%), especially during the CPB phase. Average cerebral NIRS values of patients who did not undergo circulatory arrest (CA) during CPB, 10 min after CPB weaning, were associated with average CI values with a significant correlation (r = 0.7, P = 0.003). In conclusion, during CPB, cerebral NIRS values are expected to remain constantly lower than somatic sensors, which instead tend to show similar elevated saturations, regardless of their position. Based on these results, positioning of noncerebral NIRS sensors during CPB without CA may be questioned. Key Words: Near-infrared spectroscopy—Pediatric cardiac surgery—Cardiopulmonary bypass—Deep hypothermic circulatory arrest.
Clinical utilization of near infrared spectroscopy (NIRS) has recently increased (1), several years after its first conception (2). In particular, the application of NIRS monitoring in pediatric cardiac surgery has
gained ongoing success due to its reliability as a perfusion assessment tool, ease of use, and low invasiveness (3). NIRS uses principles of light transmission and absorption to noninvasively measure the concentrations of oxygenated (HbO2) and deoxygenated (Hb) hemoglobin in tissue (4). NIRS instruments may be divided into two distinct types: concentration monitors and saturation monitors (4). In vivo saturation spectrometers are those most commonly used and described in clinical practice: they do not attempt to measure concentrations of HbO2 and Hb in tissue, but instead measure the ratio of the two
doi:10.1111/aor.12424 Received May 2014; revised September 2014. Address correspondence and reprint requests to Dr. Zaccaria Ricci, Department of Cardiology, Cardiac Surgery, Pediatric Cardiac Intensive Care Unit, Bambino Gesù Children’s Hospital, IRCCS, Piazza S. Onofrio 4, 00165 Rome, Italy. E-mail: [email protected] Artificial Organs 2015, 39(7):584–590
NIRS DURING PEDIATRIC CARDIAC SURGERY and derive the hemoglobin oxygen saturation (4). Some of these instruments, furthermore, attempt also to reduce the effects of surface tissues overlying the brain by measuring the absorption ratios at two different source-detector spacing and subtracting one from the others (5). Several targets have been proposed as objects of NIRS scanning: cerebral saturation, as a marker of brain perfusion or as a surrogate of superior vena cava oxygen saturation (ScVO2); splanchnic saturation, as a marker of abdominal perfusion; and renal saturation, as an index of renal blood flow adequacy (6). As a matter of fact, currently, NIRS monitoring reliability and application during pediatric cardiac surgery, both in the intraoperative and postoperative phases, are debated (7). In particular, the role of multisite sensor positioning and monitoring has not extensively been evaluated, and its application is limited by the cost of each disposable sensor pad. Cardiopulmonary bypass (CPB) is one of the most delicate phases of open-heart pediatric surgery, due to the risk of cerebral and systemic malperfusion. Although several technical aspects recently developed for pediatric CPB have significantly improved the outcome of children with congenital heart disease after cardiac surgery (8), multiorgan protection during the nonpulsatile phase of CPB is still a primary target. There is still great uncertainty about the management of most important clinical parameters, such as optimal temperature, hematocrit, blood flow, perfusion pressure, metabolic strategy, etc. (8–10). Multisite monitoring of perfusion during CPB could contribute to address this important issue. The aim of our study was to evaluate absolute values of multisite NIRS monitoring during CPB in a cohort of infants undergoing pediatric cardiac surgery to describe average differences between cerebral, renal, upper body (arm), and lower body (thigh) regional saturation. Furthermore, the correlation between cerebral NIRS and cardiac index (CI) at CPB weaning was evaluated. PATIENTS AND METHODS Design A retrospective analysis of prospectively collected data was conducted on infants undergoing pediatric cardiac surgery with CPB with a four-channel NIRS intraoperative recording. The INVOS 5100C Cerebral Oximeter (Somanetics, Troy, MI, USA) was used in all patients. From June to December 2013, to evaluate institutional standardization of intraoperative monitoring, all infants were provided with multisite NIRS after anesthesia induction: four NIRS
585
probes were placed on the right forehead, on the right posterior flank, on the right upper arm, and on the right thigh to estimate the cerebral, renal, and upper and lower body muscle perfusion, respectively. Neonatal probes were used in patients weighing less than 5 kg, and pediatric probes were applied in all other patients. Recorded data were retrieved from the INVOS monitor and utilized for the analysis. The INVOS monitor automatically recorded two data points per minute in all monitored patients. All patients with a number of missing data points exceeding 10% of the overall monitored time were not included in the analysis. Inclusion and exclusion criteria and data collection Inclusion criteria were infants with elective surgery, in stable hemodynamic condition, and younger than 1 year old. Demographic, surgical, and clinical data (Aristotle score, vasoactive inotropic score [VIS], Pediatric Index of Mortality 2 [PIM2], length of mechanical ventilation [MV], pediatric cardiac intensive care unit length of stay [PCICU LOS]) were obtained from an institutional database, and CPB data were retrieved from perfusionists’ electronic database. General anesthetic and CPB management All patients received general anesthesia with sevoflurane at induction and midazolam (0.05 mg/kg/ h), fentanyl (5 μg/kg/h), and rocuronium (0.5 mg/ kg/h) infusion for maintenance. CPB was always primed with packed red blood cells (PRBCs) and 5% albumin, and conducted with alpha stat strategy and blood cardioplegia. A CPB flow of 150 mL/kg/min, a hematocrit of 28–30%, and moderate hypothermia with the temperature ranging between 25 and 32°C were targeted. Deep hypothermic circulatory arrest (DHCA) at 18°C was used when necessary. The standard institutional protocol for CPB weaning consisted of the use of milrinone (0.75 μg/kg/min and dopamine 5 μg/kg/min) started at the end of crossclamp as first choice. If MAP was below 45 mm Hg and filling pressures did not indicate the need for fluid replacement, dopamine was increased to 10 μg/ kg/min. If such targets were not achieved, adrenaline (0.05–0.3 μg/kg/min) was added. Evaluated scores and hemodynamic measurements Aristotle score is currently used in our department to classify surgical risk of our patients (11). VIS was utilized to assess patients’ hemodynamic status at CPB weaning and it was calculated as follows: dopamine dose (μg/kg/min) + dobutamine dose (μg/kg/min) + 100 × epinephrine dose (μg/kg/min) + Artif Organs, Vol. 39, No. 7, 2015
586
Z. RICCI ET AL.
10 × milrinone dose (μg/kg/ min) + 10 000 × Vasopressin dose (units/kg/min) + 100 × norepinephrine dose (μg/kg/min) (12). PIM2 score is utilized in our hospital to assess critical illness severity of children admitted to our pediatric intensive care units (PICUs) (13). Finally, after CPB, CI is routinely monitored at our institution by pressure recording analytical method (PRAM) (14): it is a minimally invasive hemodynamic monitoring system based on mathematical analysis of the invasive arterial waveform recorded at a high sampling rate (1000 Hz). The area under the pressure wave is examined during the whole cardiac cycle, including the postdicrotic notch phase, and it is utilized to assess the patients’ stroke index. PRAM monitoring requires an arterial waveform without artifacts and the correct identification of the dicrotic notch (it cannot be used during CPB and in case of under or overdamping); PRAM monitor automatically records each minute’s CI and makes it available for download. The institutional review board of the Bambino Gesù Children’s Hospital approved the protocol and waived the need for informed consent due to the retrospective nature of the study [Prot n. 235LB]. Study objectives The primary aim of this study was to describe the relative average value of the four regional saturations as measured by NIRS sensors during the main intraoperative phases: from anesthesia induction (after intubation) to sternotomy (baseline), from sternotomy to CPB start (T1), from CPB start to CPB weaning (T2), and from CPB weaning to sternal closure (T3). We also attempted to evaluate if significant differences were present between values recorded by each sensor, during each intraoperative period. Third, we compared the NIRS values in the subgroup of patients undergoing CPB without DHCA (noCA) with those requiring DHCA (CA). In patients undergoing DHCA, this was assessed as an adjunctive intraoperative phase (T2CA) interposed between the phase before (T2pre) and after the arrest (T2post). Finally, we analyzed the reliability of cerebral NIRS as indicator of global perfusion, as surrogate of ScVO2, during CPB weaning by correlating it with CI values: the average cerebral NIRS value and the average CI value of each patient during the first 10 min after CPB weaning were considered. Statistical analysis The Mann–Whitney test or Student’s t-test were used, according to the variable distribution, to compare continuous variables. Two-way analysis of variance (ANOVA) for repeated measures was used Artif Organs, Vol. 39, No. 7, 2015
to assess NIRS changes over the four predefined time points. To assess the association between regional saturation (continuous variable) and other covariates (intraoperative phases [T1, T2, T3, and TCA] and measurement sites), a generalized estimating equation (GEE) multivariable linear regression model was conceived, considering the nature of the data. This choice was made for the ability of GEE to model the dependence structure of the data. In this analysis, an unstructured working correlation has been used. The model also included, in addition to main effects, interactions between the channel and measurement sites. The normality of the residuals was tested to assess the adequacy of the model. Spearman’s correlation was used to verify univariate association between continuous variables. Data are presented as median and interquartile range (IQR) or average and standard deviation (SD) or standard error (SE) where necessary. A P value 0.05, except where indicated by the symbols. * P = 0.028; ** P = 0.044. † Preoperatively. X-Clamp, cross-clamp.
represents the tissue oxygen average saturations at the four sites in the noCA and CA groups: two-way ANOVA showed that time (P < 0.0001 in both groups) and site (P = 0.0001 in the noCA group and P = 0.012 in the CA group) significantly affected variations of NIRS values. Once cerebral NIRS values were excluded from the two-way ANOVA, the impact of sensor site was no longer significant (P = 0.184 in the noCA group and P = 0.42 in the CA group). For this reason, cerebral NIRS was used as reference for channel variable in the GEE linear regression model: it was associated with a significantly different saturation level, at intercept, than that of other channels (always P < 0.0001), with similar estimates between arm (+10.36 [0.6]), flank (+5.88 [1.13]), or thigh (+10.69 [1.05]) saturations (Table 2). Detailed analysis of interaction between site of NIRS measurement and time point showed that cerebral NIRS was always significantly lower than that of other channels (P < 0.0001). Furthermore, at T2 NIRS saturation increased at all sites: cerebral NIRS saturation rise according to GEE was significantly lower with respect to somatic values (P < 0.0001). Finally, during the phase of CA, all saturations were lowest (although still significantly different between cerebral and others) (Table 2).
Hemodynamic monitoring after CPB weaning When average cerebral NIRS values of noCA patients after CPB weaning (T3) were associated
FIG. 1. Average modifications of NIRS over time. Panel A: patients without circulatory arrest (noCA). Panel B: patients undergoing CA. Data are depicted as mean and standard deviation. In panel B, T2CA represents the CA phase and it is interposed between the CPB phases before (T2pre) and after the arrest (T2post). Artif Organs, Vol. 39, No. 7, 2015
588
Z. RICCI ET AL. TABLE 2. GEE of NIRS values at different sites (CH) and time points (T) Analysis of GEE parameter estimates Empirical standard error estimates Parameter Intercept T1 × CH T1 × CH T1 × CH T1 × CH T2 × CH T2 × CH T2 × CH T2 × CH T3 × CH T3 × CH T3 × CH T3 × CH T2CA × CH T2CA × CH T2CA × CH T2CA × CH
2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1
Estimate
SE
83.97 2.55 5.02 2.13 0.00 6.75 13.63 4.67 0.00 9.01 10.10 2.29 0.00 −6.44 −6.25 −6.02 0.00
4.23 0.90 1.06 0.88 0.00 0.64 1.20 1.10 0.00 0.67 1.31 1.15 0.00 1.18 1.36 1.34 0.00
95% conf limits 75.67 0.77 2.93 0.40 0.00 5.49 11.28 2.51 0.00 7.69 7.52 0.03 0.00 −8.77 −8.93 −8.66 0.00
92.27 4.33 7.12 3.86 0.00 8.01 15.99 6.84 0.00 10.33 12.68 4.55 0.00 −4.11 −3.57 −3.39 0.00
Z
P
19.82 2.81 4.70 2.42 — 10.48 11.35 4.24 — 13.36 7.68 1.99 — −5.42 −4.57 −4.49 —
