To the Editor:
We have read with interest the article entitled “Elevated cerebral
pressure passivity is associated with prematurity-related intracranial
hemorrhage” by O’Leary et al in the recent issue of the journal (1). The
authors’ previous findings suggest that, in preterm neonates during
postnatal transition, cerebral pressure passivity cannot be predicted by
blood pressure values alone and that it is not an “all or nothing”
phenomenon but one that presents with varying severity (2,3). In these
studies, the authors also investigated whether the prevalence and severity
of cerebral pressure passivity are related to intracranial hemorrhage
(IVH) or parenchymal echodensities in preterm neonates during the
immediate postnatal period but couldn’t arrive to a firm conclusion (2,3).
To further investigate this question, the authors in the present study
have focused on studying the suspected association between the severity
(magnitude) of cerebral pressure passivity and IVH and/or parenchymal
echodensities in 88 preterm neonates <32 weeks’ gestation. In addition
to other clinical and hemodynamic parameters, they continuously recorded
mean arterial blood pressure (MABP) and the difference (HbD) between
oxygenated hemoglobin (HbO2) and hemoglobin at 2 Hz using continuous near-
infrared spectroscopy (NIRS). The data were collected for up to 5 days,
and the authors performed coherence and transfer function analysis between
MABP and HbD signals in 3 frequency bands (0.05– 0.25, 0.25– 0.5, and
0.5–1.0 Hz). Using MABP-HbD gain and clinical variables, they then created
a logistic regression model to best predict the likelihood of the
development of IVH and parenchymal echodensities. The authors found that
low-frequency MABP-HbD gain was significantly associated with early IVH
but not parenchymal echodensities. They concluded that the
cerebrovascular monitoring technique used in this study allows
quantification of cerebral pressure passivity as MAP-HbD gain in preterm
neonates during postnatal transition. They also noted that the temporal
and causal relationship between MABP-HbD gain and IVH remains to be
determined in this patient population.
The findings of this study are interesting and clinically relevant.
As mentioned earlier, the authors reported finding no significant
correlation of IVH with the high and medium frequency gain. However, we
have concerns related to the way the frequency domain analysis was
performed, as we believe the sampling theorem was not correctly applied
due to oversimplification. In addition, the MABP signal, filtered by the
algorithm providing the signal, was given without describing the
characteristics of averaging used by the manufacturer.
As for the frequency domain analysis, the authors state, “Given our
sampling frequency of 2 Hz the Nyquist Theorem (4) allowed frequency-
domain analyses up to 1 Hz”. The sampling theorem as described by Lathi
states that “a real signal whose signal is band-limited to B Hz[X(w)=0 f
or/w/ >2piB] can be reconstructed exactly (without any error) from its
samples taken uniformly at a rate fs>2B samples per second." However,
the Nyquist-Shannon Sampling theorem is often oversimplified, as it is
believed that ’sampling at twice the maximum frequency of interest is
sufficient‘. This statement, however, ignores the requirement for the
signal to be bandlimited (i.e. of infinite duration). The fact that the
signals in the paper by O’Leary et al1 are of finite duration (non-
bandlimited) leads to a phenomenon called “aliasing” (4). Aliased data
contain ambiguous frequency data that makes analysis very difficult; this
is strikingly demonstrated by the apparent reversal of rotation of a
spinning wagon wheel once their rotation exceeds a critical frequency.
Because of the likelihood of aliasing when sampling at twice the maximum
frequency of interest, we suggest two alternatives: either choose a higher
sampling rate, filter the data, and then analyze them up to 1 Hz, or keep
the 2 Hz sampling rate, filter the data and analyze them up to a lower
frequency only.
This issue is particularly important because the authors found that
their results were significant only in the low-frequency band, and their
subsequent discussion relates only to these low frequency measures of the
MABP-HbD gain. In fact, aliasing would least affect the low frequency
components of the MABP and HbD signals. Therefore, the failure to remove
aliasing may have contributed to the lack of significant findings at
medium and high frequencies.
As for the MABP signal, our concern is that, by definition, the MABP
signal has been subjected to an averaging filter, which will change its
frequency domain characteristics. For example, it is plausible that MABP
is calculated with a running average filter, which is the worst scenario
for data to be analyzed in the frequency domain (4). If this were the
case, it would be incorrect to record the averaged arterial pressure
signal at 2Hz and then attempt to analyze it up to 1Hz without taking into
consideration the processing that was performed on it. Although the
authors may have known the characteristics of the averaging used by the
manufacturer, the article does not contain this information. Here again,
as the filtering provided by a MABP algorithm would likely least affect
the low frequency components of the MABP and HbD signals, this may have
also contributed to the lack of significant findings at medium and high
frequencies. In summary, it would have been advisable to provide the
original averaging period for the MABP data and consider this information
when reporting the lack of significance of the association between the
magnitude of cerebral pressure passivity and IVH and/or parenchymal
echodensities in medium and high frequencies.
We would like to emphasize that our criticism only implies that the
results of the study by O’Leary et al(1) need to be very carefully
interpreted and, as the authors suggested, no definitive conclusions be
drawn. Research on developmental hemodynamics in high-risk neonates
requires a thorough understanding of developmental cardiovascular
physiology, pathophysiology and outcomes research and knowledgeable
implication of appropriate technologies to monitor the relevant signals
and real-time data collection. Although we believe that these requirements
are met in the study1 discussed in our letter, we would like to draw
attention to the fact that in complex clinical research where different
disciplines such as medicine, biostatistics and mathematics merge and work
together to allow for the most appropriate analysis of the data, clarity
in the interpretation of the findings transcending disciplines must be
ensured. This can be aided by novel approaches to training
multidisciplinary researchers in medicine forming a truly integrated
research team whose members have sufficient knowledge in each discipline
to ensure appropriate interpretation of the clinically relevant findings
of these studies.
Respectfully,
Matt Borzage, MS, BME* and
Istvan Seri, MD, PhD**
Center for Fetal and Neonatal Medicine
USC Division of Neonatal Medicine
Department of Pediatrics
Childrens Hospital Los Angeles and
LAC+USC Medical Center
Keck School of Medicine
University of Southern California
Los Angeles, CA
* Ph.D. Student, Department of Biomedical Engineering, Viterbi School
of Engineering, University of Southern California and the Center of Fetal
and Neonatal Medicine at Children’s Hospital Los Angeles, Keck School of
Medicine, University of Southern California, Lois Angeles, CA
** Corresponding author
References
1. O’Leary H, Gregas MC, Limperopoulos C, Zaretskaya I, Bassan H,
Soul JS, Di Salvo DN, du Plessis A. Elevated cerebral pressure passivity
is associated with prematurity-related intracranial hemorrhage. Pediatrics
2009; 124:302-309
2. Soul JS, Hammer PE, Tsuji M, et al. Fluctuating pressure-passivity
is common in the cerebral circulation of sick premature infants. Pediatr
Res 2007; 61:467– 473
1. Tsuji M, Saul JP, du Plessis A, et al. Cerebral intravascular
oxygenation correlates with mean arterial pressure in critically ill
premature infants. Pediatrics 2000; 106:625– 632
2. Lathi BP. Linear Signals and Systems. Carmichael, CA: Berkeley-
Cambridge Press; 1992
3. Smith SW, The Scientist & Engineer's Guide to Digital Signal
Processing, California Technical Pub., 1997
Conflict of Interest:
None declared
eLetters: ![[Read eLetters]](/icons/misc/sectionDOWN.gif){kind=icon}
