Prolonged Central Venous Desaturation Measured by Continuous Oximetry Is Associated with Adverse Outcomes in Pediatric Cardiac Surgery

• Intermittent monitoring of central venous oxygen saturation, as a surrogate for mixed venous oxygen saturation, has been used as an index of global tissue perfusion and may predict adverse events following cardiac surgery

• In this unblinded study, prolonged periods of venous desaturation on continuous oximetry were associated with increased adverse events in children with congenital heart disease who were undergoing cardiac surgery

PEDIATRIC patients undergoing cardiac surgery are at risk for adverse clinical outcomes in the perioperative period because of inadequate tissue perfusion. Although several clinical and laboratory markers associated with adverse clinical outcomes have been identified, the inability to readily and reliably assess global tissue perfusion and cardiac output remains a challenge in this patient group. Similar to results in adults, significant mixed venous (SvO²) desaturations seen with surgically placed oximeters have shown to be predictive of worsening neurologic function and clinical outcomes in neonatal patients undergoing palliative repair for hypoplastic left heart syndrome.1Unfortunately, mixed venous oximetry using percutaneous catheters is not feasible or practical in most children.

Central venous saturation (ScvO²), as a surrogate of SvO², can reliably be used as a marker of global tissue perfusion. Recent studies have suggested the predictive power of intermittently measured ScvO²for adverse events following cardiac surgery.2,3However, intermittent or “snapshot” measurements of ScvO²are not able to assess the duration or temporal pattern of venous desaturations, and may not allow early detection of occult venous desaturations in the absence of other hemodynamic disturbances. Therefore, the true impact of desaturations on clinical outcomes is not clearly evident. In addition, though previous studies have suggested hemodynamics variables such as blood pressure, heart rate, and venous pressures to be poor predicators of adverse outcomes,1,4these variables remain a mainstay of critical care management in pediatric patients. Therefore, further investigation of the relationships between continuous ScvO2 and the more commonly used clinical hemodynamic and biochemical markers are needed to clinically interpret the value of continuous ScvO²monitoring in pediatric patients.

Using a newly developed, percutaneously placed continuous ScvO²oximetry system for children,5we examined the association between postoperative venous desaturations and clinical outcomes in pediatric patients undergoing cardiac surgery. We hypothesized that prolonged low central venous saturations are associated with adverse clinical outcomes.

Upon receipt of approval and informed consent from The University of California, Los Angeles Institutional Review Board, Los Angeles, California, 54 high-risk pediatric patients scheduled to undergo cardiac surgery with complete repair (no residual shunting), with or without cardiopulmonary bypass, and requiring a central venous catheter were consented for prospective data acquisition. Patient exclusion criteria included contraindications for a central venous catheter placement in the right internal jugular vein and body weight less than 3 kg.

Anesthetic induction consisted of inhaled sevoflurane and nitrous oxide, supplemented with intravenous fentanyl (3–5 μg/kg) and pancuronium (0.1 mg/kg). Isoflurane (1–1.5%) and fentanyl (10–15 μg/kg) were used to maintain anesthesia during the surgery. Following tracheal intubation, a central venous fiberoptic oximetry catheter (PediaSat™ oximetry catheter; Edwards Lifesciences, Irvine, CA) was calibrated in vitro  and placed into the right internal jugular vein. Catheter size (4.5F/5 cm, 4.5F/8 cm, or 5.5F/8 cm) was determined according to patient size. Proper placement of the catheter tip at the lower portion of the superior vena cava and superior to the caval-atrial junction was verified by transesophageal echocardiography.

Following repair, patients received inotropic support as clinically indicated, including dopamine, dobutamine, epinephrine, and/or milrinone according to standard institutional protocols. Patients were transported to the Intensive Care Unit (ICU) and catheters were recalibrated on arrival in vivo  with the most recent hemoglobin and venous blood gas co-oximetry values. Catheter tip location was reconfirmed with a chest radiograph and patients were monitored for clinical outcomes and major adverse events until the end of their hospital stay. All data were prospectively recorded.

In the Operating Room (OR), ScvO², heart rate, central venous pressure, mean arterial pressure, arterial saturation, cerebral regional tissue oxygen saturation (rSO²) (INVOS® Cerebral Oximeter; Somanetics Corporation, Troy, MI), and core temperatures were recorded every minute. In the ICU, ScvO²and rSO²were recorded every minute, whereas heart rate, central venous pressure, mean arterial pressure, arterial saturation, and temperature gradients were recorded every 15 min. Data were recorded from the completion of surgical repair (if performed off cardiopulmonary bypass) or complete separation from cardiopulmonary bypass (if the procedure was performed on cardiopulmonary bypass) until 24 h after ICU admission. Data were collected until ICU discharge if patients were discharged before 24 h. In addition, ScvO²catheter measurements were compared to venous blood gas co-oximetry values of simultaneously drawn samples at predefined time points (OR baseline, prerepair, postrepair, chest closure, ICU admission, and 2, 4, 6, 8, 12, and 24 h postoperatively). Venous blood samples for co-oximetry measurement were drawn from the distal lumen of the catheter. Only those ScvO²catheter measurements with an adequate signal quality index from 1 to 3 were used. Serum lactate was measured at least every 6 h, or more frequently if indicated clinically. Serum creatinine measurements were drawn every 12 h. Complete rSO²data sets were acquired in a subset of 36 patients. For patients with bilateral cerebral rSO²sensors, the average was used for analysis. Postoperative clinical outcome data were prospectively recorded.

Major adverse events were defined as need for extracorporeal support, refractory cardiogenic shock or hypotension despite maximal medical therapy, sepsis, multi-organ dysfunction, and reoperation or reopening of chest in the first 24 h. We selected previously validated clinical outcome measures to assess the postoperative course of the study patients.6,–,8Total length of intubation, length of ICU stay, and length of hospital stay were measured in days for each patient. Creatinine clearance was calculated by the Cockcroft-Gault formula at 12 h post-ICU admission.9The maximum number of inotropes used during the study period, the number of inotropes still in use at 24 h, and inotrope score at 24 h were recorded. The inotrope score was calculated as described by Wernovsky et al.  with dosages in micrograms per kilogram per minute: dosages of dopamine + dobutamine + (epinephrine × 100) + (milrinone × 10) at 24 h post-ICU admission.10 

We investigated the frequency and temporal distribution of desaturation episodes detected by continuous ScvO². For a given threshold ScvO², we defined a desaturation episode as 5 or more consecutive min below the defined threshold, separated by 5 or more consecutive min above the threshold. These episodes were calculated for three separate time periods: from surgical repair/off cardiopulmonary bypass to OR discharge, first 12 h of ICU admission, and from 12 h following ICU admission until the end of data collection.

To relate intraoperative and postoperative ScvO²to clinical outcomes and hemodynamic or biochemical variables, we utilized the area under the curve (AUC) as a measure of degree of desaturation, where the area is calculated between a threshold ScvO²and the measured ScvO²values over time. The AUC is expressed as a product with the units of minute-percent and is determined as described below:

Y = measured ScvO²value, T = specified threshold value. We compute Z as the value of difference between the measured and specified threshold ScVO²according to the definition:

The AUC for each patient is calculated by the following formula: AUC = (ΣZ)*t, where t = total number of minutes from the end of surgical repair up to 24 h after ICU admission. The following threshold groups were selected for analysis: AUC 60–70%, AUC less than 60%, AUC less than 50%, and AUC less than 40%.

We computed the time a patient remained under each threshold or between thresholds and, for each, determined the optimal time that best discriminates between those with major adverse events and without them. The optimal time was chosen by carrying out a nonparametric receiver operating characteristic curve analysis and was chosen such that accuracy was maximized. The accuracy was defined as the average of the sensitivity and specificity (accuracy = 0.5 sensitivity + 0.5 specificity). For the optimal time we report the sensitivity, specificity, accuracy, and positive and negative predictive values, assuming that the major adverse event prevalence is the same as in our data.

The associations of ScvO²AUC with clinical outcomes, standard hemodynamic measurements, and biochemical markers (lactate) were assessed using the Spearman rank test and by fitting restricted cubic splines. The nonparametric Spearman correlation and the corresponding splines using log scale AUC was used, because the relationship is not necessarily linear, even on the semilog scale. Using the spline fit, the average change in the outcome from the 25^thto 75^thpercentile change in each log ScvO²AUC was computed. This is essentially a nonparametric, robust slope. A P  value <0.05 was considered statistically significant. Agreement between the catheter measured saturation and co-oximetry measured saturation values was assessed using Bland-Altman plot analysis.

Continuous data are presented as mean ± SD, or median with range as appropriate. Statistical analysis was performed with SPSS 16.0 (SPSS Inc., Chicago, IL) and SAS 9.2 (SAS Inc., Cary, NC).

Of the 54 patients enrolled, 3 patients were not included in the final analysis because of incomplete data sets caused by data storage errors. Another patient was excluded because of extracorporeal membrane oxygenation support being initiated in the operating room. Patient characteristics and surgical procedure for the 50 remaining patients are summarized in table 1. Given the diverse patient population, each patient's age, weight, preoperative diagnosis, and surgical procedure are provided in the appendix. No patient had a demonstrable intracardiac shunt on intra- and postoperative echocardiography regardless of undergoing single or bi-ventricular repair. Continuous ScvO²data were recorded from the end of surgical repair and up to the first 24 h postoperatively and lasted a median of 24 h (interquartile range 19.4–25.8 h). Only 2 patients were discharged from the ICU sooner than 24 h. There were no complications attributable to ScvO²monitoring or continuous data collection. Patient outcomes and clinical data are summarized in table 2.

Major adverse events were observed in 9 patients and are summarized in table 3. The sensitivity and specificity of ScvO²as a predictor for major adverse events were calculated using the time the ScvO²was less than 40%, less than 45%, less than 50%, less than 55%, less than 60%, and from 60 to 70%. The accuracy of each time was calculated as the average of sensitivity and specificity. For the time with the greatest accuracy, the positive and negative predictive values for major adverse events were determined for various venous saturation thresholds and are summarized in table 4. Using receiver operating characteristic curve analysis, it was determined that a time greater than 18 min of ScvO²less than 40% was the most predictive for major adverse events with 100% sensitivity, 97.6% specificity, 98.8% accuracy, 90% positive predicative value, and 100% negative predictive value (fig. 1). Increasing time of ScvO², such as less than 45% and less than 50%, were only slightly less predictive. A time greater than 93 min of ScvO²less than 55 percent was still highly predictive of major adverse events with an accuracy of 80.5%, a positive predictive value of 50%, and a negative predictive value of 97.1%. A ScvO²of less than 60% was only predictive of major adverse effects with a duration exceeding 13 h. ScvO²of 60–70% was not predictive of major adverse effects at any length of time. Using ScvO²AUC instead of time produced similar results.

Four ScvO²AUC groups were considered for analysis for correlation with clinical outcomes: AUC 60–70%, less than 60%, less than 50%, and less than 40%. These data are summarized in table 5. As AUC thresholds were decreased in 10% increments from AUC 60–70% to AUC less than 40%, an increasing number of associations with clinical outcomes were observed. At ScvO²AUC less than 40%, significant correlations were found between creatinine clearance at 12 h postoperatively (r = −0.58), length of ICU stay (r = 0.56), maximum number of inotropes used (r = 0.52), number of inotropes used at 24 h (r = 0.40), inotrope index score at 24 h post-ICU admission (r = 0.39), length of hospital stay (r = 0.36), and length of intubation (r = 0.32). Statistically significant, but slightly weaker, correlations resulted between ScvO²AUC less than 50% and creatinine clearance 12 h postoperatively (r = −0.56), length of ICU stay (r = 0.44), max inotrope use (r = 0.39), inotrope use at 24 h (r = 0.33), length of hospital stay (r = 0.30), and inotrope index score (r = 0.30). Significant correlations were observed only between AUC less than 60% and creatinine clearance 12 h postoperatively (r = −0.52), length of ICU stay (r = 0.33), and length of hospital stay (r = 0.32)^.ScvO²AUC 60–70% was not significantly correlated with any of the clinical outcomes.

In addition to the correlation of ScvO²AUC and clinical outcomes, the magnitude of effect of increasing AUC and the change in the outcome measure was determined between the 25^thand 75^thpercentile in each ScvO²AUC range. For a ScvO²AUC less than 40%, length of intubation increased by 1.43 days, length of ICU stay increased by 4.32 days, length of hospital stay increased by 1.93 days, maximum inotrope use increased by 1.01, inotrope use at 24 h increased by 0.74, the inotrope index score increased by 6.05 mcg/kg/min, and creatinine clearance at 12 h decreased by 24.53 ml/min. For an ScvO2 AUC less than 50%, the length of ICU stay increased by 6.01 days, length of hospital stay increased by 6.9 days, the maximum inotrope use increased by 0.76, the maximum inotrope use at 24 h increased by 0.54, the inotrope index score increased by 3.79 mcg/kg/min, and the creatinine clearance at 12 h decreased by 26.93 ml/min. The ScvO2 AUC less than 60% demonstrated an increase in the length of ICU stay and hospital stay by 4.98 and 5.44 days, respectively, and a decrease in creatinine clearance at 12 h of 15.33 ml/min.

Desaturation episodes were calculated for ScvO²thresholds of less than 40%, less than 50%, less than 60%, and less than 70%, and are summarized in table 6. The highest number of observed desaturation episodes occurred within the first 12 h after surgical repair.

Regression analysis demonstrated no significant association between ScvO²measurements and hemodynamic variables such as blood pressure, heart rate, and central venous pressure (P > 0.05) in the ICU. Core temperature and temperature difference also did not correlate with ScvO². In addition, calculated AUC for greater than 25% change in mean arterial pressure or heart rate from clinically normal values for age11were not significantly associated with any clinical outcome.

Though serum lactate was found to not correlate with ScvO²AUC at any threshold, lactate levels proved to be an independent predictor of poor outcomes, demonstrating an association with creatinine clearance 12 h post-ICU admission (r = −0.4), length of ICU stay (r = 0.4), length of hospital stay (r = 0.4), inotrope use at 24 h (r = 0.4), inotrope index score (r = 0.4), and max inotrope use (r = 0.3).

In the subset of 36 patients in whom rSO²AUC was recorded in addition to ScvO²AUC, rSO²and ScvO²were found to be significantly correlated in the OR (r = 0.51) and ICU (r = 0.3). rSO²AUC less than 40% was significantly correlated with creatinine clearance (r = −0.48) and length of ICU stay (r = 0.40). Fewer significant associations were seen with rSO²AUC less than 50%, which correlated with only length of ICU stay (r = 0.39).

The accuracy and precision of the pediatric ScvO²oximetry system was monitored throughout the study. Bland-Altman plot analysis of 341 paired data sets of venous blood gases and ScvO²catheter simultaneous measurements showed difference of means (bias) and precision to be 0.41%± 4.93%, respectively.

Our investigation demonstrates that low perioperative ScvO²is associated with poor clinical outcomes in pediatric patients undergoing cardiac surgery. This is the first study to demonstrate that pediatric patients with a larger area under the curve of continuous ScvO²saturation below 40% had increased length of intubation, stay in the ICU, stay in the hospital and inotrope use, along with lower creatinine clearance in the perioperative period. An AUC less than 50% proved to be almost equally associated with poor clinical outcomes. An AUC less than 60% had weaker association with poor clinical outcomes, whereas AUC 60–70% had no significant associations.

Furthermore, ScvO²was also an excellent predictor of major adverse effects. Our results from receiver operating characteristic curve analysis demonstrate that periods as brief as 18 min below saturation of 40% are highly predictive of major adverse effects. We observed that venous saturations of less than 50% for greater than 34 min are also significantly predictive of major adverse effects. The data also show that more mild venous desaturations are well tolerated for significantly longer periods of time, without increasing the risk of major adverse effects. In addition, ScvO²values at various thresholds maintained a strong negative predictive value, demonstrating that avoidance of lower venous saturations is a valid goal in pediatric patients undergoing cardiac surgery to decrease the risk of major adverse events.

Our findings indicate that the largest incidence of significant ScvO²desaturations predominantly occur within the first 12 h postoperatively. During these periods where ScvO²fluctuations are commonly anticipated, continuous ScvO²monitoring, rather than intermittent venous sampling, would be extremely valuable for detection of venous desaturations to prompt earlier interventions. Finally, ScvO²with fiberoptic oximetry was accurate with a low bias and good precision (0.41%± 4.93%), supporting the potential use of continuous ScvO²as a reliable target parameter in high-risk pediatric patients.

The clinical value of ScvO²monitoring was originally proposed in patients with cardiac disease in 1970.12Adult clinical trials have suggested that low ScvO²saturations are associated with increased risk of mortality and independently associated with poor clinical outcomes. Suggested target ScvO²values from these studies ranged from 60 to 73%.13,–,22Surprisingly limited such data exists for pediatric congenital heart disease patients, and clinical practice is often dictated by experience from adult studies. Our results suggest that increasingly poor clinical outcomes are seen with ScvO²less than 60%, with significantly worse outcomes present with ScvO²less than 40%. However, poor correlations were found between continuous ScvO²and commonly measured hemodynamic variables, suggesting that ScvO²better reflects global cardiopulmonary changes and tissue perfusion. This observation is also supported by previous experimental findings demonstrating that decreases in ScvO²can reflect tissue hypoxia despite apparently normal hemodynamic parameters.1,5 

While no study to date has investigated the relationship of lactate elevation resulting from anaerobic conditions and continuous ScvO², several pediatric studies have found that lactate was an independent predictor of poor clinical outcomes.2,23–24Our results agree with these findings in that higher lactate measurements were associated with longer stays in the ICU and hospital, higher inotrope use, and lower creatinine clearance postoperatively. It is not unexpected that of the other measured physiologic parameters, ScvO²would be most correlated with rSO², because the two parameters are affected by similar physiologic principles governing oxygen delivery. However, correlations in the ICU were much weaker when compared with those in the OR. In the 35 patients with complete rSO²data sets, significant associations were also observed between rSO²AUC and clinical outcomes; these associations were not as significant as those seen with ScvO². This difference may be due to a more pronounced influence of pCO²on rSO²values, whereas ScvO²values are largely independent of ventilation. In part, the weaker association between rSO²and clinical outcomes may also be because of the assumptions incorporated into the algorithms used in the monitoring technology (site of measurement in forebrain), or because of the variability incurred as a result of altering sensor contact on the forehead when placed for extended periods. This also may be a statistical phenomenon because there were fewer patients with complete rSO²data sets.

Continuous fiberoptic venous oximetry in pediatric patients has recently been shown to be accurate and poses minimal additional risks compared with standard central venous catheter placement.5,25In the presence of tissue hypoxia, continuous monitoring of ScvO²may allow for detection of decreased tissue oxygen delivery earlier in comparison to traditional, and typically late, indicators of organ dysfunction, such as increased lactate, decreasing urine output, and increasing metabolic acidosis. Real-time information about tissue oxygenation in pediatric patients can have a profound impact in decreasing morbidity by detecting tissue hypoperfusion earlier than other hemodynamic and chemical parameters. In addition, the use of a real-time perioperative ScvO²monitor is particularly useful in pediatric patients to decrease the number of blood draws required for venous saturation monitoring. Since the monitoring of continuous ScvO²saturations is accurate, uses a parameter familiar to critical care physicians, and has no further risk to the patient than standard central venous catheter placement, it would seem a logical step to add this monitor to standard perioperative care in high-risk pediatric patients. Clinical utilization of continuous ScvO²monitoring should continue to take into account the complexity of the patient's surgical repair, other hemodynamic parameters, and biochemical monitors to make a comprehensive assessment.

While the findings of our investigation are promising, we acknowledge several limitations. Our study was an observational, unblinded prospective trial, and although poor clinical outcomes and major adverse events strongly correlated with ScvO²less than 40% and 50%, a prospective trial to evaluate the utility of continuous ScvO²as a target parameter of goal-directed therapy to improve clinical outcomes is warranted. Technical limitations of the ScvO²catheter were minimal from our experience and included wall artifacts caused by impingement of the catheter tip on a vessel wall that led to an unreliable signal, as detected by a signal quality index value of 4. In all cases, repositioning of the catheter resulted in restoration of an appropriate signal. Although the study was not blinded, the practice in our ICU before the study and the study protocol required therapeutic interventions to be made primarily using confirmed venous desaturations from intermittent venous blood gas samples and/or other clinical indicators. There is some possibility that bias might have been introduced with the presence of the monitor and that the information could have been used to guide therapy. The availability of continuous venous oximetry may have influenced some of the clinical outcomes, such as escalation of inotropes, delayed extubation, or delayed discharge. However, the unblinded nature of the study should not have influenced the occurrence of major adverse events or parameters such as creatinine clearance. In addition, there was no treatment algorithm provided for the treatment of low continuous ScvO²and the physicians involved in the study were not obligated to use the continuous ScvO²values to guide therapy. The fact that the information was available to the treating physicians, but that significant adverse outcomes were still observed, may be demonstrative that even more aggressive intervention needs to be undertaken to resolve low venous saturations in a timely fashion.

In summary, our prospective observational study demonstrates a significant association between intraoperative and postoperative ScvO²desaturations and unfavorable clinical outcomes in pediatric patients. Further investigation should focus on quantifying ScvO²temporal responsiveness to changes in standard hemodynamic parameters and defining target ScvO²levels specific to patient groups as part of a blinded clinical trial.

The authors thank Jeffery Gornbein, Ph.D., Professor, Department of Biostatistics, University of California-Los Angeles, Los Angeles, California, for performing the statistical analysis used in the study and Jennifer Scovotti, M.A., and Cecilia Canales M.S. (Research Associates), Department of Anesthesiology, University of California-Los Angeles, for helping with the data collection.