Near-infrared spectroscopy has been used to monitor cerebral oxygen saturation during cerebral circulatory arrest and carotid clamping. However, its utility has not been demonstrated in more complex situations, such as in patients with head injuries. The authors tested this method during conditions that may alter the arteriovenous partition of cerebral blood in different ways.
The authors compared changes in measured cerebral oxygen saturation and other hemodynamic parameters, including jugular venous oxygen saturation, in nine patients with severe closed head injury during manipulation of arterial carbon dioxide partial pressure and after mean arterial pressure was altered by vasopressors.
The Bland and Altman representation of cerebral oxygen saturation versus jugular oxygen saturation showed a uniform scatter. Values for changing arterial carbon dioxide partial pressure were: bias = 1.1%, 2 SD = +/-21%, absolute value; and those for alterations in mean arterial pressure: bias = 3.7%, 2 SD = +/-24%, absolute value. However, a Bland and Altman plot of changes in cerebral oxygen saturation versus changes in jugular oxygen saturation had a negative slope (alteration in arterial carbon dioxide partial pressure: bias = 2.4%, 2 SD = +/-17%, absolute value; alteration in mean arterial pressure: bias = -4.9%, 2 SD = +/-31%, absolute value). Regression analysis showed that changes in cerebral oxygen saturation were positively correlated with changes in jugular venous oxygen saturation during the carbon dioxide challenge, whereas correlation was negative during the arterial pressure challenge.
Cerebral oxygen saturation assessed by near-infrared spectroscopy does not adequately reflect changes in jugular venous oxygen saturation in patients with severe head injury. Changes in arteriovenous partitioning, infrared-spectroscopy contamination by extracerebral signal, algorithm errors, and dissimilar tissue sampling may explain these findings.
NEAR-INFRARED spectroscopy (NIRS) is a method for continuous, noninvasive monitoring of cerebral oxygen saturation (ScO2). Its accuracy depends on distinguishing the signal reflected by the brain from that reflected by other tissues (skin, muscle, bone). 1,2Several studies have shown that when used in normal human subjects, NIRS responds rapidly to cerebral oxygen desaturation during marked cerebral hypoperfusion or systemic hypoxia. 3–5However, because severe hypotension (e.g. , circulatory arrest) influences both cerebral and extracerebral compartments, such studies do not validate the technique or at least do not prove its value in more complex situations. For example, NIRS detects only a fraction of the pathologic events identified by parallel multimodal monitoring in human head injury. 6Moreover, when NIRS is used in an intensive care setting, changes in cerebral blood flow (CBF) or vascular tone may modify the arteriovenous (AV) distribution of blood in the brain differently from the distribution in scalp and in other extracerebral tissues. All of these phenomena may considerably distort ScO2. The present study therefore examines the relations between NIRS ScO2and jugular venous oxygen saturation (SvjO2) during changes in arterial carbon dioxide partial pressure (PaCO2) and blood pressure in adults with head trauma. Our working hypothesis was that these tests produced different effects on CBF and AV partition, thus identifying some of the limitations of NIRS in clinical use.
Materials and Methods
This study was approved by the ethics committee of the hôpital H. Mondor, Créteil, France, and informed consent was obtained from patients’ next of kin. Nine patients with severe closed head injury (Glasgow Coma Scale < 8) and with multifocal contusions or diffuse brain swelling confirmed by computed tomography scan were studied within the first 10 days after injury. Patients with bilateral frontal contusions were not included. In order to be able to measure all signals on the same side, patients who had focal contusions on the side of the dominant jugular vein were also excluded. All patients were placed in the supine position with the head and thorax tilted upward at 30°. They were sedated with midazolam and fentanyl and were ventilated mechanically to achieve 100% arterial oxygen saturation (SaO2) and moderate hypocapnia (PaCO230–35 mmHg). Patients who had a mean arterial pressure (MAP) < 70 mmHg were given a norepinephrine infusion to maintain arterial pressure above this threshold before the start of the study. All patients had normal core temperature and blood hemoglobin level > 10 g/100 ml.
A retrograde jugular catheter (Opticath 5.5 French; Abbott, Rungis, France) was placed on the side corresponding to the dominant jugular vein, as assessed by intracranial pressure (ICP) increase during a compression test. 7The position of the catheter in the jugular bulb was confirmed by radiograph. ScO2was recorded by NIRS (Invos 3100; Somanetics, Troy, MI). This system used two wavelengths: 730 and 810 nm. The sensor contained a near-infrared light-emitting diode, and two light detectors located 30 and 40 mm from the light-emitting diode, to distinguish between the cerebral and extracerebral signals and thus monitor the ScO2in the underlying area of the brain. The sensor was placed on the upper forehead, on healthy skin, 4–5 cm from the midline, ipsilateral to the jugular catheter. Background light was excluded by an opaque cap. The blood flow velocity in the middle cerebral artery (MCAv) was recorded unilaterally by pulsed Doppler ultrasound (Angiodyne; DMS, Montpellier, France), which was placed on the same side as the NIRS probe. The jugular catheter was always ipsilateral to both ScO2and Doppler probes.
The following parameters were recorded at 200 Hz (Labview 4.0; National Instrument, Austin, TX) and averaged every 4 sec: ICP (ventricular catheter), MAP (radial catheter), end-tidal CO2(PETCO2; Merlin, Hewlett Packard, Les Ulis, France), MCAv, and ScO2. All signals were stored for off-line analysis. SvjO2was determined by sampling blood via the jugular catheter.
Jugular blood samples were taken and ScO2was simultaneously recorded at three levels of PETCO2:(1) during moderate hyperventilation (PaCO2: 30–35 mmHg, T0);(2) during intense hyperventilation (PaCO2): 20–25 mmHg, T1); and (3) during a short period of moderate hypercapnia induced by CO2inhalation (T2). CO2inhalation was achieved by connecting the air inlet of the ventilator to a chamber (AGA, Toulouse, France) containing 5% CO2in 95% O2. The inspiratory CO2could thus be changed by altering the setting on the ventilator for fraction of inspired oxygen without altering the rate of ventilation or the tidal volume. SvjO2was not allowed to decrease to < 50% during hyperventilation, and ICP was not allowed to increase to > 35 mmHg during CO2inhalation. To standardize the procedure, the order of changes in PaCO2was the same in all patients.
Jugular blood samples were taken in the basal state (T0), during hypocapnia (T1), and during hypercapnia (T2). Steady state was maintained for at least 2 min (range, 2–10 min) at T0and T1. The rapid increase in ICP at T2made it impossible to reach steady state. All patients underwent two or three CO2challenges, with an interval of at least 24 h between each.
An equivalent of cerebral vascular resistance (EqCVRI) was estimated from the ratio of cerebral perfusion pressure (CPP) to MCAv as proposed by Aaslid et al. 8Changes in each variable studied during CO2challenge corresponded to the difference between the reading after the test minus that measured before the test.
Changes in Systemic Arterial Pressure
A formal test of the response to change in arterial pressure was performed once in each patient and was started after 20 min of steady state after the end of the first CO2challenge. Therefore, the systemic arterial pressure was pharmacologically modified under the same level of moderate hypocapnia as that achieved at T0. This was performed once in each patient by starting or increasing the rate of norepinephrine infusion (n = 4) or by discontinuing or reducing in the rate of norepinephrine perfusion (n = 5). Increasing or starting norepinephrine was used in patients whose MAP corresponded to values close to the lower limit of the autoregulatory plateau. Conversely, we stopped or reduced norepinephrine infusion in patients whose MAP was close to the upper limit of the autoregulatory plateau. In all patients, ICP was continuously monitored, and norepinephrine infusion was stopped or restarted (as appropriate) if ICP reached 35 mmHg. We also continuously monitored SvjO2via the fiberoptic catheter to ensure that it did not decrease to < 50% during the test. Arterial and jugular blood samples were taken when a stable condition was achieved after a step change in norepinephrine.
Results are expressed as means ± SD. Differences between the means of the parameters determined during the CO2challenge, at T0, T1, and T2were compared by two-way analysis of variance for repeated measurements. The Bland and Altman test 9was used to compare the absolute values of ScO2and SvjO2obtained from jugular blood samples. This test was also applied to changes in ScO2(δScO2) and changes in SvjO2(δScO2). The relative δScO2for T0–T1and T0–T2, were compared with the corresponding δSvjO2by linear regression. Data of the pressure test were analyzed in the same way as for the CO2challenge.
Carbon Dioxide and Arterial Pressure Challenges
The data (means ± SD) obtained during CO2challenge and during changes in MAP are shown in tables 1 and 2. Data of ScO2and SvjO2were plotted according to Bland and Altman. 9We saw a similar scatter of error for both tests (fig. 1): the mean of individual differences (ScO2− SvjO2) for CO2and arterial pressure challenges were 1.1 and 3.7. Two SDs of these individual differences amounted to ±21%(absolute value) and ±24%(absolute value) for the CO2and arterial pressure challenges, respectively. Individual differences between δScO2and δSvjO2are shown in figure 2. There was an apparent systematic decrease of individual differences between δScO2and δSvjO2when (δScO2+δSvjO2)/2 increased. This seems to reflect a bias that is not randomly distributed.
To better understand this effect we performed a linear correlation between δScO2and δSvjO2for both tests (fig. 3). The slopes were both significantly different from unity (P < 0.05, with an SE of the slope = 0.033 and 0.11 for the CO2and the pressure challenge, respectively). They were positive in the CO2challenge but negative in the arterial pressure challenge. Hence, regression shows, in addition to the Bland and Altman analysis, that both tests were not only different in terms of the magnitude of the response but the relationship between δScO2and δSvjO2during CO2and pressure were opposite.
The main finding of our study is that ScO2assessed by NIRS does not adequately reflect changes in SvjO2in patients with severe head injury. Unlike “conventional” hypoxic challenges, we chose to examine effects that could result in changes in the AV partition of cerebral blood volume in a possible contamination by extracerebral signals, and by different tissue fields being monitored. Patients with head injuries were studied rather than normal volunteers. This setting introduced more complexity because any regional difference in cerebral perfusion regimen could increase the discrepancy between SvjO2and ScO2.
Arterial Pressure Challenge
An arterial pressure challenge in normal patients should not alter CBF because of autoregulation. SvjO2should also not vary, as cerebral metabolic rate in oxygen (CMRO2) was likely to be constant throughout the study. But autoregulatory vasoconstriction 10,11could reduce arteriolar blood volume. Venous blood volume should be unaffected if CBF remains constant. Thus, an increase in MAP in a context of effective autoregulation should lead to a decrease in ScO2at constant SvjO2. However, autoregulation is often impaired in patients with head injuries, as in our patients. Indeed, we observed that SvjO2and MCAv decreased with decreasing MAP during the hypotensive challenge, despite a 17% reduction in EqCVRI. Conversely, SvjO2and MCAv increased during norepinephrine infusion, and EqCVRIalso increased by 50%. Hence, in this condition of partially altered autoregulation, arteriolar vasoconstriction was insufficient to increase cerebral vascular resistance proportionally to cerebral perfusion pressure, and this was accompanied by an increase in CBF. This condition should have led to a decrease in the capillary oxygen gradient, a greater capillary contribution to the arterial component of the ScO2signal, but also to an increase in SvjO2and venous blood volume. The opposite phenomena was expected during vasodilation. We observed that δScO2and δSvjO2correlated negatively when arterial pressure was changed. This tends to support the fact that changes in AV partition should have occurred with a larger increase in the venous compartment than the increase observed in the arteriolar one.
Carbon Dioxide Challenge
The decrease in cerebral vascular resistance in response to vasodilatation of the pial arterioles at T2was reflected in a 34% decrease in EqCVRIwithout change in MAP. Vasodilation should have allowed the volume of the arteriolar compartment to increase. The increase in blood flow at constant cerebral metabolic rate in oxygen should result in an independent increase in the oxygen saturation of the venous compartment and a passive increase in its volume. 12Hence, hypercapnia, like a change in arterial pressure, may alter the AV partition. However, in contrast to the arterial pressure challenge, we observed a positive correlation between δScO2and δSvjO2(fig. 3). Thus, the change in AV partition probably occurred during a CO2challenge different from that observed during an MAP change. One must also consider possible differential effects on extracerebral tissues. Moreover, ScO2, which includes arteriolar blood, would be expected to be higher than SvjO2. This was only true at T2in our study. All of these issues deserve comment.
First, SvjO2could be “artifactually high” compared with ScO2because it incorporated deep brain structures that extract less oxygen than the neocortex structures monitored by NIRS. Second, the cerebral signal could be contaminated by a reflected signal from extracerebral structures (e.g. , bone, muscle) with unpredictable partition and O2saturation characteristics.
This contamination could account for most of the discrepancies observed between tests. The suggestion that the musculocutaneous territory makes a major contribution to NIRS signals is supported by recent experimental data, which indicate that the CBF measured by NIRS is three times greater when the probe is placed on the dura than when it is measured through the scalp. 13Similarly, theoretical and experimental investigations of the propagation of light through the skull have shown that light is poorly absorbed by brain grey matter, as it accounts for only 15% of the total light absorbed. 14
The impact of extracerebral tissues on ScO2recording is critical. CO2dilates cerebral vessels, as well as those in the musculocutaneous territories such as the forehead and scalp. 15There should therefore be a parallel between the changes in blood flow in both territories when PaCO2is altered. The infusion of norepinephrine when autoregulation is disturbed should lead to a decrease in blood flow in the forehead and scalp and an increase in CBF as MAP increases. Although we did not measure velocity at the site of ScO2measurement or in extracerebral tissues, we postulate that opposite responses by extracerebral tissues during changes in CO2and MAP, combined with changes in AV partition, could account for our findings. Such circumstances would not occur in studies based on hypoxic challenge because oxygen desaturation occurs simultaneously in all territories during such hypoxic experiments, including the musculocutaneous territory. 3The same criticism can be made about measurements conducted after severe hypotension during ventricular fibrillation. 5The inability of studies to cope with complex causes of inaccuracy may explain the persisting questions about the use of NIRS in clinical medicine.
Moreover, the physiology of our patients was not simple. Regional vascular variations may occur in patients with head injuries, with hypoperfused ischemic areas and hyperemic areas, leading to nonuniform oxygen extraction and venous saturation. 16CBF–CO2reactivity and autoregulation may also be patchy. 17,18Some investigators suggest that that these may account for the discrepancies between changes in ScO2and SvjO2. 19,20Metz et al. 21showed, by bilateral SvjO2measurement in patients with head trauma, that unilateral SvjO2measurement did not reliably reflect ischemic events. In contrast, the difference in SvjO2between each side did not change with time. Tateishi et al. 19reached the same conclusion after reporting a fair correlation between δSvjO2and δScO2during changes in PaCO2in patients with acute brain disease. This suggests that interhemispheric inhomogeneity may not have had much effect in our study, which considered variations in SvjO2and in ScO2. However, there may have been some intrahemispheric difference, especially between the anterior cerebral artery territory above which the NIRS sensor was placed and the middle cerebral artery territory, which was monitored by Doppler ultrasound. There is no reason to believe that this difference should be larger than the interhemispheric difference.
Apart from the problems originating from AV partitioning and the role of extracerebral tissue previously discussed, there may be other errors in the NIRS. The optical path length and light scattering may vary. 22The technology we used includes a continuous wave and, thus, cannot correct for these sources of error. The limits of this method may also have been reached in our study, where different challenges may have induced differing effects on optical path length and light scattering in addition to the possible influences of the AV partition and extracerebral tissues.
The authors thank Proffesor B. Riou, Department of Anesthesiology, Centre Hospitalier Universitaire Pitié-Salpétrière, Paris, France, for discussions of the manuscript.