Authors' Reply

To the Editor:

We thank Wolf and colleagues for their comments concerning our article (Newton et al., 1997) and address their comments.

The complete statement, quoted from our manuscript, was “In conclusion, NIRS detects blood flow in the brain in most subjects, but measurement of CBF_NIRS in the reflectance mode, using oxygen as an intravascular tracer is inaccurate, particularly at high CBF values.” We were concerned not only about bias and linearity, but also the large test-retest variability encountered in the study. Variability of the magnitude encountered in our study is unacceptable for a clinical measurement. However, the authors do raise interesting questions regarding the source of this variability.

What is the primary source of error in the O₂ CBF_NIRS technique, as applied to adults, children, and larger animal models? Is it attributable to the mode used to acquire information about the tracer (hemoglobin); the influence of the layers, through which light must pass, on it's information content; or is it technical error attributable to an inability to adequately characterize the information contained in the Hb and HbO₂ transients within a single transit time? We believe all of these factors are involved.

The fact that our results correlate better at lower flow rates compared to higher flow rates regardless of geometrical considerations indicates a problem specific to short transit times of tracer through the brain. We agree with Wolf and colleagues that faster sampling with the same or better signal to noise would be an improvement, though this was limited by the NIRS instrument used in our study (NIRO 500, Hamamatsu Photonics, Hamamatsu, Japan). However, a more important factor at short transit times is the tracer-input function which, with the O₂ method, is limited by the speed with which alveolar PO₂ can be increased. Pulmonary and hemodynamic problems may constrain the range over which both the magnitude and rate of increase of HbO₂ can be forced. Spontaneous changes in ventilation can exert secondary influences on flow, independent of oxygen, via carbon dioxide, and blood pressure may be unstable in some patients. At adult CBF values, it may be necessary to use bolus injections of indocyanine green to obtain rapid tracer increases in 1 to 2 seconds, as shown recently by Roberts et al., 1998).

In addition, we showed that using O₂ as a tracer does lead to changes in CBF itself as measured by total cerebral venous outflow. The observed variability of this effect is likely to result in increased test-retest variability and bias. Again, using an inert tracer such as indocyanine green would reduce this problem.

The terms “reflectance mode” and “transmission mode” are relevant but do not fully explain all the important anatomical differences between our animal study, as a model of an adult or child study, and the previous human neonate validations. In the newborn study the head is smaller, and the optodes tend to be 1–2 cm further apart. Hence, the angle between source and detector is nearer “transmission” in the newborn and nearer “reflectance” in the adult. Other important characteristics are that the extracerebral tissues (skin and skull) are significantly thinner and more transparent (less scattering) and the white matter is less scattering in the newborn compared to the human adult, human child, and adult dog. All these factors lead to better sampling of the gray matter relative to extracerebral tissues in the newborn. Perhaps it would have been better to use the terms “adult” and “neonatal” mode instead of “reflectance” and “transmission“. The major effect of these anatomical differences is their affect on the bias of the NIRS technique compared to microspheres. Of greater concern is the effect that the clear CSF layer might have on the measurement which is significantly less predictable than the other issues (Okada and Delpy, 1996; Okada et al, 1997). It is reasonable to say that potentially, channeling of light in the CSF layer will be more problematic in the adult than the neonate.

The criteria for accepting or rejecting blood flow measurements are rigorous. It should be pointed out that most of the criteria are focused on eliminating confounds of physiologic origin, rather than operator origin. Even the most highly skilled operator is not always aware of subtle changes until it is time to compute the measurement. We found that approximately two-thirds of the measurements did not fulfill established acceptance criteria. Hence, we were left with the best possible measurement conditions, and even these measurements had an unacceptably high variance when compared with a standard technique.

So how do we proceed at this juncture? What validation tests need to be performed? Test-retest procedures evaluate technique-instrument internal consistency. Clearly, reproducibility is an important validation issue; but test-retest procedures have no direct bearing on either the accuracy or bias of a measured variable. Unfortunately, the only way we have of evaluating techniques to measure CBF is by comparison against established or standard techniques. Xenon clearance techniques have been widely used to measure CBF in humans, but are not without measurement error. Intra-arterial xenon has both accuracy and a low measurement variability; but has fallen into disuse because of it's invasive nature. Both inhaled xenon and intravenously injected xenon techniques suffer from confounding components attributable to nasal passages and airways. All of the gaseous clearance techniques are most accurate at low blood flow rates and become biased at high flow rates because of gas partition problems. The microsphere technique is a recognized standard for CBF measurement in animal models, but cannot be used in humans because of the need for tissue sampling. Whereas it does report a measurement with a range of variability, statistically, it is the most accurate and least-biased technique at high blood flow rates. When we compared the O₂ CBF_NIRS measurement with this standard, using an animal model, the O₂ CBF_NIRS was found to have an unacceptable range of accuracy. Whereas O₂ CBF_NIRS generally trended blood flow, the variance associated with the measurement was too large to attribute adequate confidence to it.

We stand by our original conclusion regarding the inaccuracy of the O₂ CBF_NIRS method at blood flow rates and transit times in the adult range and our data clearly show this point. We do not dispute the findings of higher accuracy in the validations studies in newborns, although we would still like to see test-retest variance improved further. The recent results using indocyanine green and NIRS (Roberts et al., 1998) seem promising up to the maximum flow tested (50 mL 100 g⁻¹ min⁻¹). However, this technique will have to be adapted to a less invasive method to gain more widespread use.