Near-Infrared Spectroscopy Cerebral Oxygen Saturation Thresholds for Hypoxia–Ischemia in Piglets

MATERIALS AND METHODS

Near-infrared spectroscopy

By monitoring light-intensity signals at several wavelengths where oxyhemoglobin and deoxyhemoglobin have different absorption coefficients, it is possible to determine oxyhemoglobin and deoxyhemoglobin concentrations through an expression based on the principles of the Beer Law,

∑ 1 n log ⁡ ( I o / I ) = ∑ 1 n ( ε L C )

(1)

where I_o and I are the emitted and detected light intensities, L represents the path length of light through the tissue, C the concentration of the compound, e the extinction coefficient of the compound, and n is the number of wavelengths of the light being used. In the brain, the main absorbing compounds in the near-infrared spectrum are oxyhemoglobin and deoxyhemoglobin and, to a lesser extent, water and cytochrome aa₃ (Wray et al., 1988). The Sco₂ is given by

S c o 2 = ( H b O 2 ) / ( H b O 2 + H b )

(2)

where Hbo₂ and Hb represent oxyhemoglobin and deoxyhemoglobin concentrations. The Sco₂ is calculated from Eqs. 1 and 2 for three wavelengths to account for oxyhemoglobin and deoxyhemoglobin and the other absorbing compounds (Kurth and Thayer, 1999).

At present, NIRS instrumentation is based on continuous-wave, spatial-domain, frequency-domain, or time-domain technologies (Chance, 1998). In frequency-domain devices, the emitted intensity is oscillated at high frequency. After passing through the tissue, the light is amplified and demodulated. Oxyhemoglobin, deoxyhemoglobin, and Sco₂ are determined from the amplitude and phase signals, which account for I_o, I, and L in Eq. 1.

In this study, the frequency-domain NIRS consisted of a main unit, emitter, and detector fiberoptic bundles, and a computer. The main unit uses laser diodes at wavelengths of 754, 780, and 830 nm. The laser intensities are oscillated at 120 MHz. The fiberoptic bundles deliver the light to and from the head. The main unit uses homodyne frequency-domain technology to determine amplitude and phase at each wavelength. The computer captures these signals and calculates Sco₂ in real time by means of an algorithm. Instrument Sco₂ drift is 1%/h and precision is +5% (Kurth and Thayer, 1999).

RESULTS

Table 1 displays the physiologic data for the Sco₂ groups. Baseline Sco₂ for all animals was 68 ± 5%. In the Sco₂ 80% or greater group, all animals received hyperoxic hypercarbic ventilation to achieve the targeted Sco₂. In the Sco₂ 60% to 80% group, all animals continued at baseline (normoxic, normocarbic ventilation). All animals in the Sco₂ 50% to 59% group received carotid arterial occlusion. Of the animals in the Sco₂ 40% to 49% group, six (60%) received only carotid arterial occlusion, and four (40%) received both carotid arterial occlusion and hypoxic gas ventilation. All animals in the Sco₂ groups 10% to 19%, 20% to 29%, and 30% to 39% received both carotid arterial occlusion and hypoxic gas ventilation to achieve the targeted Sco₂. Technical problems precluded obtaining sagittal sinus blood in 24 piglets.

OPEN IN VIEWER

TABLE 1.

Physiologic data for the cerebral oxygen saturation groups

	Sco2 (%)
	≥ 80 (n = 4)	60–79 (n = 16)	50–59 (n = 8)	40–49 (n = 10)	30–39 (n = 10)	20–29 (n = 9)	<20 (n = 3)
pH	7.13 ± 0.08	7.41 ± 0.07	7.46 ± 0.08	7.38 ± 0.07	7.32 ± 0.15	7.38 ± 0.07	7.35 ± 0.07
Pco2 (mm Hg)	73 ± 15	36 ± 6	33 ± 4	38 ± 6	38 ± 7	33 ± 4	33 ± 6
Po2 (mm Hg)	515 ± 56	85 ± 12	88 ± 8	63 ± 20	38 ± 12	30 ±5	27 ± 10
MAP (mm Hg)	65 ± 7	61 ± 7	82 ±5	71 ± 13	59 ± 11	53 ± 9	62 ± 7
Hb (g/dL)	9 ± 1	9 ± 1	9 ± 1	9 ± 1	9 ± 1	9 ± 1	10 ± 1
Sco2 (%)	88 ± 5	70 ± 5	54 ± 3	44 ±2	35 ±2	26 ±2	18 ± 1
Svo2 (%)	82 ± 6	49 ± 7	34 ± 4	20 ± 10	5 ± 1	5 ± 2	4 ± 1
CBF (%)	147 ± 20	100	74 ± 22	50 ± 17	33 ± 20	23 ± 15	19 ± 11
Lactate (μmol/g)	5.4 ± 0.4	5.1 ± 5.0	8.4 ± 5.4	19 ± 17	58 ± 38	61 ± 36	64 ± 53
ATP (μmol/g)	2.05 ± 0.01	1.94 ± 0.08	1.85 ± 0.13	1.88 ± 0.14	1.79 ± 0.28	1.47 ± 0.14	1.56 ± 0.23

Mean ± SD.

Sco₂, cerebral oxygen saturation; Hb, arterial hemoglobin concentration; MAP, mean arterial pressure; Svo₂, sagittal sinus oxygen saturation; CBF, cerebral blood flow (by percentage of baseline); ATP, adenosine triphosphate. pH, Pco₂, and Po₂ refer to arterial blood values.

Figures 1 to 4 illustrate statistically significant relations between Sco₂ and Svo₂, CBF, and brain tissue lactate and ATP concentrations. Linear relations were observed between Sco₂ and Svo₂ (Fig. 1) and CBF (Fig. 2). Of note, the y-intercepts between Sco₂ versus Svo₂ and Sco₂ versus CBF were significantly less than zero. Curvilinear relations were observed between brain tissue lactate concentration (Fig. 3) and brain tissue ATP concentrations (Fig. 4). For lactate, as Sco₂ decreased below 40% to 45%, lactate concentration increased, suggesting a change in cerebral metabolism below this Sco₂. For ATP, as Sco₂ decreased below 30% to 35%, ATP concentration decreased, suggesting a change in cerebral energy state below this value of Sco₂.

OPEN IN VIEWER

FIG. 1.

Relation between near-infrared spectroscopy Sco₂ and Svo₂ in piglets.

OPEN IN VIEWER

FIG. 2.

Relation between near-infrared spectroscopy Sco₂ and CBF in piglets.

OPEN IN VIEWER

FIG. 3.

Relation between near-infrared spectroscopy Sco₂ and brain tissue lactate concentration in piglets.

OPEN IN VIEWER

FIG. 4.

Relation between near-infrared spectroscopy Sco₂ and brain tissue ATP concentration in piglets.

Figures 5 and 6 and Table 2 depict logistic regressions to define changes in cerebral function in terms of EEG activity, abnormal tissue lactate, and ATP concentrations for the population of animals. Normal brain tissue lactate and ATP concentrations were, respectively, 0 to 15 and 1.78 to 2.10 μmol/g tissue (Table 1, 95% confidence intervals for Sco₂ 60% to 79% group). As Sco₂ decreased below 50%, minor EEG changes and abnormally high lactate levels were observed in some animals. As Sco₂ decreased below 40%, all animals had minor EEG changes and abnormally high lactate levels, and some animals also had major EEG changes and abnormally low ATP levels. As Sco₂ decreased below 30%, all animals had major EEG changes and abnormally low ATP levels.

OPEN IN VIEWER

TABLE 2.

Logistic regression coefficients

	X50	K	Max	Min	r2
Lactate	44.0	0.294	101	−0.74	0.99
ATP	33.8	0.524	98	−0.74	0.99
Major EEG	37.5	0.341	100	0.02	0.99
Minor EEG	42.8	0.098	100	−13	0.97

Coefficients for the logistic regression equation: Y = min + (max - min)/(1 + e^−K*(X-X50)).

ATP, adenosine triphosphate; EEG, electroencephalography.

OPEN IN VIEWER

FIG. 5.

Logistic regression plot between near-infrared spectroscopy Sco₂ and the percent of piglets without minor change in EEG activity or without major change in EEG activity compared with baseline. Data points indicate mean ± SD. Logistic regression equation variables are presented in Table 2.

OPEN IN VIEWER

FIG. 6.

Logistic regression plot between near-infrared spectroscopy Sco₂ and the percent of piglets with normal brain tissue lactate concentrations and normal brain tissue ATP concentrations. Data points indicate mean ± SD. Logistic regression equation variables are presented in Table 2.

Table 3 displays the Sco₂, CBF, and Svo₂ thresholds for 50%, 5%, and 95% of animals based on the logistic regression. The thresholds for CBF and Svo₂ were calculated from the linear regressions (Figs. 1 and 2) using the Sco₂ threshold calculated from the logistic regression. Lactate increases and minor EEG changes occurred first, followed by further reduction in Sco₂, CBF, and Svo₂ levels, after which major EEG changes occurred and ATP decreased. For 50% of animals, the Sco₂ thresholds ranged from 44% (increased lactate) to 33% (decreased ATP), the Svo₂ thresholds from 23%(increased lactate) to 8% (decreased ATP), and the CBF thresholds from 56% (increased lactate) to 36% (decreased ATP) of baseline perfusion.

OPEN IN VIEWER

TABLE 3.

Cerebral hypoxia–ischemia thresholds

	Sco2 (%)	CBF (% baseline)	Svo2 (%)
Increased lactate	44 (33–52)	56 (37–70)	23 (8–33)
Minor EEG change	42 (26–70)	52 (24–100)	20 (0–57)
Major EEG change	37 (29–46)	42 (29–59)	12 (3–25)
Decreased ATP	33 (28–39)	36 (28–47)	8 (1–16)

Values where 50% (95% to 5%) of animals experience increased tissue lactate concentration, electroencephalography (EEG) changes, or decreased tissue adenosine triphosphate (ATP) concentration. N = 60 animals.

Sco₂, cerebral oxygen saturation; CBF, cerebral blood flow; Svo₂, sagittal sinus oxygen saturation.

DISCUSSION

Detection of cerebral hypoxia–ischemia remains problematic in neonates because neurologic examination findings, CBF, and Sjo₂ monitoring are unreliable or impractical (Hill and Volpe, 1999). Although NIRS, a relatively new technology, is suited for this purpose, cerebral oxygenation thresholds for hypoxia–ischemia have not been defined. In adults, CBF and Sjo₂ are used to detect cerebral ischemia, by comparing observed and threshold values at which brain function and structure become impaired (Heiss, 2000; Hossmann, 1994; Schell and Cole, 2000). In our study, increased tissue lactate, minor and major EEG changes, and decreased tissue ATP levels were observed at Sco₂ values of 44%, 42%, 37%, and 33%, respectively, indicating hypoxic–ischemic thresholds for functional impairment occur at Sco₂ values of 33% to 44%. These thresholds were well below baseline the Sco₂ at 68%, suggesting a buffer exists between normal and dysfunction, similar to that for CBF and Sjo₂.

The classic concept of viability thresholds for cerebral ischemia distinguishes between CBF thresholds for functional neurologic impairment and structural tissue damage. As CBF decreases, functional impairment occurs first and is evident almost immediately after CBF drops below the threshold; structural damage occurs after more severe reductions in CBF and is time dependent (Heiss, 2000). These CBF thresholds, defined in adult animals and humans, vary somewhat with ischemic duration, temperature, and anesthesia. (Hossmann, 1994; Heiss, 2000). Examples of functional impairment include altered brain protein synthesis, lactate concentrations, synaptic activity, ATP concentrations, and membrane potential occurring at CBF reductions of 30% to 50%, 50% to 60%, 60% to 75%, 75% to 85%, and 85% to 90%, respectively (Hossmann, 1994). For structural damage, selective neuronal death and tissue infarction occurs after CBF reductions of 60% to 75% and 80% to 95% for at least 30 minutes (Hossmann, 1994).

In our study, lactate, ATP, and EEG were used to assess functional impairment in which the changes were evident within 30 minutes after Sco₂ was decreased. Structural damage was not assessed. However, the CBF thresholds for selective neuronal death and infarction are similar to those for increased lactate and decreased ATP (Hossmann, 1994). Thus, NIRS Sco₂ thresholds for structural damage may be similar to those for functional impairment, provided the hypoxia-ischemia is of sufficient duration. The NIRS Sco₂ thresholds, like the CBF thresholds, may vary somewhat with ischemic duration, temperature, and anesthesia. For comparison, we calculated CBF thresholds from Sco₂ thresholds using the CBF and Sco₂ relation (Table 3). These calculations revealed CBF reductions associated with increased lactate, EEG changes, and decreased ATP that were in the same range as those reported in adult models of ischemia (Hossmann, 1994).

Bilateral carotid occlusion alone in piglets decreased CBF by 26% with nonsignificant changes in lactate and ATP concentrations (Table 1). Cerebral vasodilation, increased arterial pressure, and collateral blood flow through the basilar circulation and circle of Willis helped maintain CBF. With superimposed arterial hypoxia, however, arterial pressure decreased, and CBF decreased to a greater extent along with lactate and ATP concentrations. This pressure passive CBF response reflects maximal cerebral vasodilation with bilateral carotid occlusion. Arterial pressure could have fallen during hypoxia from a decrease in cardiac output or systemic vascular resistance.

In adult clinical care, Sjo₂ monitoring is often used to detect cerebral ischemia (Schell and Cole, 2000). The Sjo₂ thresholds for functional and structural neurologic impairment are not as well defined as for CBF. The Sjo₂ values in healthy volunteers range from 55% to 71% (Gibbs et al., 1942). During graded arterial hypoxia in awake human volunteers, major EEG changes occur at Sjo₂ below 40%, confusion develops at Sjo₂ below 33%, and loss of consciousness at Sjo₂ below 26% (Gibbs et al., 1942). In cerebral ischemia, increased tissue lactate and neurologic deficits can appear at Sjo₂ below 50% in humans, though tissue ATP does not decrease until Sjo₂ below 15% in cats (Schell and Cole, 2000). In our study, we calculated Svo₂ thresholds for the piglets, analogous to Sjo₂ in humans, from the Sco₂ thresholds using the Svo₂ and Sco₂ relation (Table 3). These calculations showed increased lactate, EEG changes, and decreased ATP developed at Svo₂ below the aforementioned values for Sjo₂. This finding likely reflects different experimental models.

Logistic regression analysis was used to determine the hypoxia–ischemia thresholds. This analysis determines the interaction between a quantal dependent variable (present or absent) and a continuous independent variable. In our study, the independent variable was Sco₂. The dependent variable was presence or absence of EEG change, or of abnormal lactate and ATP. The advantage of this approach is that it permits the use of all Sco₂ measurements to determine the “critical” value rather than only the low values of Sco₂. The number of animals in each Sco₂ group was set to optimize curve fitting and to permit the calculation of probability of functional impairment or brain damage, which helps in the design of clinical trials. Probabilities (thresholds) for 50%, 5%, and 95% of animals were calculated for biologic and clinical monitoring reasons. An alternative approach, reduce Sco₂ in each animal in a stepwise fashion until functional impairment occurs, is confounded by the effect of number and duration of the preceding steps.

The NIRS is a regional oxygenation monitor. Theoretical work and imaging experiments illustrate that it views a banana-shaped tissue field between emitter and detector (Chance, 1998; Okada et al., 1995), which in our study encompassed neocortical gray and white matter, the ischemic region. The NIRS Sco₂ represents a mixed vascular hemoglobin-oxygen saturation of capillaries, arterioles, and venules in that tissue field. Because of the greater volume of the venous circulation, Sco₂ is closer to venular than arterial saturation. In clinical care, the use of NIRS Sco₂ thresholds may not detect cerebral hypoxia–ischemia, if, for example, the cortex included a mixture of perfused and ischemic tissue such that the average Sco₂ was above the threshold, or if NIRS monitors frontal cortex and focal ischemia occurs in the occipital cortex.

Our study illustrates a few points about NIRS and physiology. In principle, cerebral oxygen delivery and consumption regulate Sco₂. Cerebral oxygen delivery consists of CBF, arterial oxygen saturation (Sao₂), hemoglobin concentration, and hemoglobin-oxygen affinity. It is easy to manipulate Sco₂ through CBF or Sao₂ by changing the inspired oxygen and carbon dioxide concentrations, minute ventilation, or occluding the cerebral arterial circulation. The presence of an arterial component in the Sco₂ is shown by the positive value of Sco₂ as Svo₂ approaches zero (x-intercept, Fig. 1). The slope of this line shows the dominance of the venous contribution to Sco₂, which is close to unity. The inability of brain tissue to extract all the oxygen bound to NIRS-monitored hemoglobin is shown by the positive value of Sco₂ as CBF approaches zero (x-intercept, Fig. 2).

The logistic regression analysis also shows great similarity among the relations for lactate, major EEG change, and ATP. The K values are nearly identical and the curve fits are tight (Table 2), suggesting a strong link to the same phenomena (i.e., cell energy dysfunction). In contrast, for minor EEG change, the K value is different and the curve fit is less tight, suggesting linkage to another phenomena associated with cerebral hypoxia (e.g., synaptic dysfunction).

Our results help clarify the use of NIRS hemoglobin oxygenation monitoring in clinical care. In healthy human neonates, children, and adults, Sco₂ ranges from 60% to 80%, similar to that in newborn pigs (Brun et al., 1997; Fantini et al., 1999; Kurth et al., 2001; Springett et al., 2001). The Sco₂ threshold for functional neurologic impairment begins at a level 25% less than this range, suggesting the existence of a buffer zone (Sco₂ 45% to 60%), in which cerebral oxygenation is less than normal but adequate for function. This Sco₂ difference between normal and dysfunction is well within NIRS precision (± 5%), affording some luxury in clinical monitoring. In many sick neonates, cerebral hemoglobin oxygenation is at or below the functional Sco₂ threshold, indicating that cerebral hypoxia-ischemia may not be recognized during clinical care (Brun et al., 1997; Tsuji et al., 1998; Watzman et al., 2000; Kurth et al., 2001; Fantini et al., 1999; Springett et al., 2001).

There are several types of NIRS instruments available for use that report hemoglobin oxygenation differently, such as tissue oxygenation index, oxydeoxyhemoglobin concentration difference, Sco₂, or regional oxygen saturation (Brun et al., 1997; Chance, 1998; Fantini et al., 1999; Springett et al., 2001; Tsuji et al., 1998; Watzman et al., 2000). These devices differ in the technology and algorithm used to determine hemoglobin oxygenation. If these devices measure hemoglobin saturation accurately at low saturation and in the neocortex without extracranial contamination, the ischemic thresholds for a given experimental paradigm should be similar. Given these caveats, our results should be generally applicable to the interpretation of cerebral oxygenation by all NIRS instruments.