Article Review: The Causes of Cerebral Desaturation in
Article Review: The Causes of Cerebral Desaturation in
The study by Brinkman, Amadeo, Funk, and Grocott (2013) investigated cerebral desaturation, which frequently occurs in one-lung ventilation procedures for thoracic surgery and is possibly caused by lower cardiac output. The aim of the study was to explore the correlation between decreases in cerebral oxygen saturation that occurs in patients undergoing one-lung ventilation and decreased cardiac output. Although cerebral oxygen saturation was lowered in all of the participants, the researchers did not have statistically significant data that could be used to establish a correlation between cerebral desaturation and cardiac output, so the hypothesis had to be rejected.
Purpose
The main purpose of the study was to establish a correlation between cardiac output and cerebral oxygen desaturation. Various hemodynamic variables, including mean arterial pressure, stroke index, stroke volume variability, and cardiac index, were measured. In addition to cerebral oxygen saturation, peripheral oxygen saturation was measured.
Although the correlation between cerebral desaturation and cardiac output was not established, it is possible to conclude that cerebral oxygen desaturation occurs in all patients who undergo one-lung ventilation, and that finding has an important implication for anesthesiologists. When anesthetizing patients, it is important to monitor cerebral oxygenation to protect the patients’ brain tissue.
However, even for non-anesthetists, the study can be useful. According to Brinkman et al. (2013), cerebral oxygen desaturation is universal in patients during one-lung ventilation, and anesthesiologists used sevoflurane to maintain anesthesia on the fraction of inspired oxygen (FiO2) of 1.0. In similar studies, all patients showed decrease in mini-mental state examination scores, and decrease in cerebral oxygen saturation levels was also correlated with various cognitive problems (Brinkman et al., 2013). Those findings can be implemented in post-operative treatment regimes for patients because to help them resolve possible cognitive issues and ensure a better recovery.
Because the study implies that patients can potentially experience cerebral ischemia, it is important to consider the subsequent supportive care for patients who display symptoms of cerebral ischemia. For example, systemic blood pressure needs to be maintained and nutrition requirements should be designed to prevent hyperglycemia.
Most importantly, the intraoperative decrease in cerebral oxygen saturation of 12 percent or more was associated with increased patient morbidity in carotid endarterectomy procedures (Brinkman et al., 2013). While the role of the anesthesiologist is to deliver oxygen to the patients when necessary, surgeons and policy-makers are responsible for determining which variables need to be monitored and which protocols need to be followed under different circumstances. For example, the study did not find any correlation between peripheral and cerebral oxygen saturation, so decreased cerebral oxygen saturation with low FiO2 could go undetected if the proper variables are not monitored and increase the risk of non-pulmonary damage and morbidity rates. According to Brinkman et al. (2013), current regulations in one-lung ventilation management requires the administration of lower FiO2 if peripheral saturation is at acceptable levels. However, it is evident that there is no relationship between the two variables, so cerebral hypoxia could be induced with current regulations. It is important to monitor cerebral oxygen saturation separately and regulate FiO2 values based on those values rather than peripheral saturation measurements.
The study is also beneficial for researchers who want to investigate the mechanisms of cerebral oxygen desaturation in one-lung ventilation procedures further. Although the study was underpowered, the researchers reported a significant relationship between decreased heart rate and decreased cerebral oxygen saturation. Also, there was no significant correlation between peripheral saturation and cerebral desaturation or arterial pressure and cerebral desaturation. However, the research does give further pointers for future studies because investigating other variables, such as cerebral blood flow velocity, is required for gaining a better understanding of the cerebral desaturation mechanism.
It is also important to note that the study was underpowered, so it is not possible to make solid conclusions based on the data, but it is also not possible to discard the potential of the research design. The same study with a larger sample might have produced different results and findings, so repeating the study may contribute to a better understanding of the variables that could influence the decrease in cerebral oxygen saturation.
Hypotheses
It was suggested that cardiac output decreases upon increase in pulmonary resistance during lung collapse when the patient is under anesthesia, which blunts the compensatory reflex mechanisms. With impaired right ventricular performance, it is possible that the right-sided filling pressures would cause increase in cerebral venous blood volume, thus affecting cerebral oxygen saturation. According to possible explanations of the mechanisms behind the decrease of cerebral saturation, Brinkman et al. (2013) hypothesized that cerebral oxygen desaturation was caused by the reduced cardiac output because it would compromise oxygen flow to the brain during one-lung ventilation.
In addition, it was hypothesized that cerebral oxygen desaturation occurs in patients undergoing one-lunge ventilation. Although previous research already supported that hypothesis, it was not supported by enough researchers to become a theory, so the authors were required to support it in their research sample to investigate the causality of cerebral desaturation further. The hypothesis was supported because a significant decrease in cerebral saturation was observed in one-lung ventilation.
However, the main hypothesis of the research was rejected because of the small sample size and lack of statistical significance when observing the relationship between the independent variable and the dependent variables. With the data obtained from measuring the cardiac output, the linear regression showed that cardiac index was unaltered in one-lung ventilation while heart rate decreased and stroke index increased. The finding was interpreted as a reflex reaction in response to lung collapse.
Because the relationship between heart rate and cerebral oxygen saturation was the only statistically significant relationship, it is not possible to conclude that decreased cardiac output is the cause of cerebral desaturation. If the hypothesis was supported, the researchers would have found that reduced cardiac output was the cause for cerebral oxygen desaturation and would have been able to clarify that mechanism. By finding the cause, they also could have discussed clinical implications of their findings and make recommendations for future research that would focus on manipulating cardiac output for increasing cerebral oxygen saturation.
Sample Size and Power
The sample size of 18 people proved to be too small and could not be used to make a conclusion. According to Brinkman et al. (2013), the study is underpowered because the relationship between the integral of cerebral desaturation over time and the increased integral of cardiac index over time was not statistically significant (P = .24). To identify a significant relationship between the two variables (P < .05), a sample size 47 patients would be required under the assumption that the R-squared level measured in the correlation maintained the same value.
All studies that fail to prove the statistical significance of their variables are considered underpowered because they most likely have lower sample size or population variability. If the threshold for statistical significance is set at P < .05, that means the results reflect a confidence interval of 95 percent. However, because the results showed high probability (P = .24), that means the confidence interval is 76 percent, so it is possible to conclude that the sample size is too small and cannot be used to accurately generalize the findings from the sample to the general population. That would result in a large sampling error, which means the sample mean value and the population mean value would significantly differ.
The R-squared value is the coefficient of determination that can have any value between 0 and 1. If the R-squared value is closer to 1, it means the multiple regression equation is good in terms of predicting outcomes. It is often used to test whether or not the independent variables individually influence the dependent variable significantly. Because the cardiac index vs. cerebral oxygen saturation linear regression analysis showed R-squared = .08, it is possible to conclude that the model and its predicting abilities are not particularly useful for the study, and the statistical significance of the correlation was too low. If a larger sample size was used, it would have been possible to improve the statistical significance of the relationship, even with a low R-squared value. For the experiment to be valid, all of the results need to pass a significance test.
The population represented by the sample consists of elderly people with a mean age of 65 years. The standard deviation was 11 years. The study investigated cerebral oxygen desaturation in both male and female patients, so 44.4 percent of them were male and 55.6 percent were female participants. Most of the patients can be considered overweight because the mean body mass index (BMI) was measured at 27 with a standard deviation of 5.
Statistical Analysis
The repeated measures design was used because it allows the researchers to compare the same variables from different groups in different conditions. For example, the hemodynamic variables, peripheral oxygen saturation, and cerebral oxygen saturation were measured during room air saturation, two-lung ventilation, one-lung ventilation, and re-inflation. A repeated measures model allows the researchers to compare those variables, even though they were measured under different conditions.
The mean length of one-lung ventilation was 101 minutes with a standard deviation of 33 minutes, so the study used mixed effects repeated measure models to assess the significance of time trends in the experiment. The mixed effect model was used because it allows continuous dependent and independent variables and can be used to investigate the interactions between continuous variables. The effects of time on hemodynamics and cerebral saturation are shown in Figure 1, and the effects of time on peripheral saturation and cerebral oxygen saturation are shown in Figure 2.
Figure 1. Hemodynamics over time. ASctO2 = cerebral oxygen saturation; MAP = mean arterial pressure (mmHg); HR = heart rate (beats*min-1); SI = stroke index (mL*m-2); SVV = stroke volume variability (%); CI = cardiac index (L*min-1*m-2). Data shown as mean (standard error). (Brinkman et al., 2013)
Figure 2. Pulse oximetry and cerebral saturation over time. SpO2 = peripheral saturation (%); ASctO2 = cerebral oxygen saturation (%). Data shown as mean (standard error). (Brinkman et al., 2013).
In that analysis, it is possible to notice that time is an independent variable that is controlled by the procedures used in the experiment. Room air, two-lung ventilation, one-lung ventilation, and re-inflation are defined as the most important points in time for reporting dependent variables, so the mean values of dependent variables for each of those time periods are reported. Hemodynamics (heart rate, arterial pressure, stroke index, stroke volume variability, and cardiac index), peripheral saturation, and cerebral saturation are dependent variables.
In the mixed effects model, fixed and random effects are differentiated, and that can be useful to this particular study. Because the primary focus was on the effects of cardiac output on cerebral oxygen saturation in one-lung ventilation, cardiac output, cerebral saturation, and one-lung ventilation can be considered fixed effects. The researchers’ interest in those variables is higher, and they would be used again when the experiment is repeated. Other hemodynamic variables and conditions can be considered random effects because their relevance in contrast to the fixed effects is low for the purpose of this study.
The post-hoc analysis was used to compare baseline and integrated changes in cerebral saturation, heart rate, cardiac index, stroke volume variability, and stroke index. The highest cerebral saturation on 100 percent oxygen was established as the baseline measurement of average cerebral oxygen saturation. The changes in cerebral saturation were monitored during the procedure and indexed to the baseline measurement. The baseline and integrated changes for heart rate, cardiac index, and stroke index were derived in a similar manner.
The purpose of post-hoc analyses is to reduce the possibility of errors that occur because of relying strictly on a priori defined statistical methods by limiting the probability of discovering false correlations. In the study by Brinkman et al. (2013), P values and confidence limits were adjusted with a simulation approach because the post-hoc comparisons had to be adjusted for multiplicity. By stating that the adjusted P values were used for post-hoc comparisons, the researchers understand that P values change during multiple comparisons among several means or groups of means and need to be adjusted because they are dependent on the entire family of comparisons.
For example, the mean values were calculated for cerebral saturation, heart rate, cardiac index, and stroke index. Because those four means need to be compared to each other, multiple comparisons are required, and the familywise significance level needs to be applied to the entire group of comparisons. Although a transformation of means can result in higher probability, a transformation of means is usually performed during descriptive statistics and was not reported in the paper. Furthermore, a transformation of means alone does not mean the P value is adjusted. For example, even after a transformation of means, if the researcher performs a single comparison, the resulting P value is still not referred to as an adjusted P value. In multiple comparisons, the P value is always adjusted to prevent type I errors.
In multiple comparisons testing, each comparison is assigned a unique adjusted P value, and those values are computed from all comparisons used in the analysis. That means the adjusted P value is dependent on entire family, and any change in the data or number of comparisons would impact the statistical significance of the comparisons.
Regular P values are used to determine the statistical significance of comparisons, but they are valid only when making individual comparisons and need to be corrected for multiple comparisons. Adjusted P values show statistical significance for each of the comparisons that have been included as part of the multiple comparison testing. If adjusted P values had not been used in this study, the researchers would have probably encountered a type I error.
Linear regression was an important statistical method used in the experiment because the goal was to establish a correlation between reduced heart rate and reduced cerebral oxygen saturation. The study also explored the correlation between other hemodynamic variables and reduced cerebral oxygen saturation. Unlike the analysis of the effect of time on dependent variables, cerebral saturation was used as an independent variable while heart rate, stroke index, and cardiac index were dependent variables because the aim of the study was to measure the impact of cerebral desaturation on factors that determine cardiac output.
Because the dependent variables (heart rate, stroke index, and cardiac index) were continuous, the researchers used a linear regression model with mixed effects. Linear regression analysis was used to compare normalized integrated changes for heart rate vs. cerebral oxygen saturation, stroke index vs. cerebral oxygen saturation, and cardiac index vs. cerebral oxygen saturation. Linear regression with mixed effects can be used to compare continuous independent variables and dependent variables, so the independent variable (cerebral oxygen saturation) in the analysis was also continuous.
In all of the analyses performed in the experiment, the independent variables can be considered within-subject variables. Within-subject variables are independent variables that are measured several times in each of the participants at different levels or time intervals. For example, one measurement of cerebral oxygen saturation is reported when breathing room air while another measurement is reported for one-lung ventilation. Because all participants were in one group that went through the same conditions, and because the type of surgical procedure was not reported in the analyses results, it is possible to assume that the research design did not include between-subject variables.
Normalized Heart Rate vs. Normalized Cerebral Oxygen Saturation
In the comparison of normalized integrated changes, the inverse correlation was established between heart rate and cerebral oxygen saturation because the linear regression’s P value was = .02. The Figure 1A, which was shown in the study, is shown here in Figure 3. It shows the results for the linear regression analysis of heart rate and cerebral oxygen saturation. The y-axis shows values for the dependent variable (normalized heart rate) while the x-axis shows the values for the independent variable (normalized cerebral oxygen saturation).
Figure 3. Normalized integrated changes in heart rate vs. cerebral oxygen saturation. ASctO2 = cerebral oxygen saturation; HR = heart rate (Brinkman et al., 2013).
Because an inverse relationship is present, a decrease in one variable’s values leads to an increase in the other variable’s values. In this case, Brinkman et al. (2013) report that heart rate decreases as the magnitude of cerebral desaturation increases. That result supports the hypothesis that dropping heart rate is correlated to the magnitude of cerebral oxygen desaturation.
However, it does not support the hypothesis that overall cardiac output is the cause of cerebral oxygen desaturation because there was no significant relationship between cerebral desaturation magnitude and stroke index or cerebral desaturation magnitude and cardiac index. Finally, even though there is an association between cerebral desaturation magnitude and decreased heart rate, it is important to take in account that cerebral desaturation is the independent variable.
Although it is possible to support the hypothesis that the magnitude of cerebral desaturation can induce cerebral ischemia by reducing heart rate, it is not possible prove that decreased heart rate is the cause of cerebral desaturation because heart rate was the dependent variable. Therefore, it is evident how heart rate responds to different cerebral oxygen saturation levels, but it is not clear how a decrease in heart rate would impact cerebral oxygen saturation.
Baseline Cerebral Oxygen Saturation
According to Brinkman et al. (2013), the cerebral oxygen saturation baseline was established by measuring “the maximal cerebral saturation on two-lung ventilation in the lateral decubitus position within ten minutes prior to lung collapse” (p. 4). Although the choice of baseline might have influenced the finding that cerebral oxygen desaturation is universal in one-lung-ventilation, the researchers believe baseline is important for the study.
First, it allows them to accurately assess the single intervention of the lung collapse. Second, various alterations in saturation levels can be determined by including the baseline. Third, when researchers agree upon a baseline, it is possible to compare and contrast the results and findings of different studies accurately. According to Brinkman et al. (2013), without a baseline, the magnitude of integrated changes varies dramatically. In fact, they suggest that animal models may be required to completely investigate the mechanisms involved in cerebral oxygen saturation decreases during one-lung ventilation.
I agree with their arguments. Proper methods and measurements ensure the quality of the study and contribute to the growing body of research. Without a universal system for measuring variables, the research results may vary significantly. Rather than contributing to research on the topic, the lack of variable control would make it impossible to contrast and discuss clinical implications by reviewing literature on the topic. Baselines should be required in all research papers to make sure that all researchers are exploring relevant variables and designing experiments in a way others can replicate them.
Even though the authors suggest that the presence of the baseline could have influenced their finding of cerebral oxygen desaturation as universal in all patients, the data and findings can be interpreted effectively only when they can be compared to other values, and baseline values need to be used for that purpose.
Conclusion
The study was underpowered, and the researchers failed to find a correlation between the decreased cardiac output and cerebral oxygen saturation. However, Brinkman et al. (2013) did report several findings that can be considered significant. First, the observation implies that prolonged cerebral desaturation and its magnitude can induce cerebral ischemia. Second, the study failed to find a correlation between peripheral and cerebral oxygen saturation. With that in mind, several strategies need to be in place to promote patient safety, such as monitoring cerebral desaturation rather than relying on peripheral saturation.
Finally, it is possible to notice that while heart rate decreased during cerebral desaturation, stroke index increased, so it is possible to suggest that a reflex mechanism is responsible for that reaction. That finding opens other possibilities for future research that will aim to clarify the mechanism of oxygen desaturation and its impact on the cardiovascular system. By investigating the cardiac reflex mechanisms further, it will be possible to repeat the study by taking in account the overlooked reflex variables and perhaps obtain different results that could impact the findings of the study significantly.
References
Brinkman, R., Amadeo, R. J., Funk, D. J., & Grocott, H. P. (2013). Cerebral oxygen desaturation during one-lung ventilation: Correlation with hemodynamic variables. Canadian Journal of Anesthesia/Journal Canadien d'Anesthésie, 1-7. doi:10.1007/s12630-013-9954-2
