Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
toring. Presently, such a study was carried out on Kolmogorov-Smirnof’s test. An estimated ICP from
the postulation that increased ICP is paralleled by PI was determined from the regression equation.
an increase in PI. The validity of the TCD(PI) method for ICP measur-
ing was tested by a Bland & Altman plot [3]. Sensi-
tivity and specificity for ICP(PI) were calculated in
Materials and Methods the ICP range of 0 to 20 and 0 to 40 mm Hg.
PATIENT POPULATION
Eighty-one patients were included with the follow-
ing types of intracranial disorders, verified by com- Results
puterized tomography (CT): 46 (57%) had aneurys- Eighty-one patients (37 males and 44 females) aged
mal subarachnoid hemorrhage, 21 (26%) severe 2 to 79 years (mean 52 years) were investigated
head injury, 7 (9%) spontaneous intracerebral hem- daily with a total of 658 PI measurements. The pa-
orrhage, and 7 (9%) had other neurologic disorders tients had a mean (⫾ SD) arterial partial pressure of
[encephalitis (n ⫽ 1), meningitis (n ⫽ 3), meningi- carbon dioxide (PCO2) of 4.84 ⫾ 0.64 (range 3.25–
oma (n ⫽ 1), idiopathic intracranial hypertension (n 7.5) kPa, a heart frequency of 79 ⫾ 17 (range 47–
⫽ 1), and sagittal sinus thrombosis (n ⫽ 1)]. 142), 97.3% with regular sinus rhythm, and a mean
INTRACRANIAL PRESSURE arterial blood pressure (MAP) of 92 ⫾ 16 mm Hg
MONITORING (range 48 –162) at the time of investigation.
The correlation between the cerebral perfusion
All patients received an intraventricular catheter
pressure (CPP) and PI was significant (p ⬍ 0.0001)
for continuous recording of the intracranial pres-
with a correlation coefficient of ⫺0.493 (Figure 1).
sure (HanniKath, 7F, pvb Medizintechnik Gmbh &
This finding was paralleled by a similar significant
Co. kg, Kirchseeon, Germany).
correlation (p ⬍ 0.0001) between CPP and ICP
TCD where the similar correlation coefficient was
One investigator conducted all TCD examinations. ⫺0.477.
TCD measurements were performed parallel to the In the overall results there was a significant cor-
ICP registration. relation (p ⬍ 0.0001) between the ICP and the PI
The daily TCD measurements were conducted with a correlation coefficient of 0.938 (Figure 2).
transtemporally using a traditional 2-MHz trans- When separating severely elevated (⬎120 cm/s)
ducer (EME TC-64 Eden medical records, Uberlin- and subnormal (⬍50 cm/s) TCD mFV values, the
gen, Germany). The TCD measurements were rou- correlation coefficient between ICP and PI was 0.828
tinely performed bilaterally on the middle cerebral (p ⬍ 0.002) and 0.942 (p ⬍ 0.638 because of a low
artery (MCA). Recordings were documented on a number of measurements), respectively (Figures 3
videographic printer for later analysis (Sony VP and 4). Only in one of 658 measurements (0.15%),
850). The depth and angle of insonation giving the the mFV exceeded 120 cm/s in conjunction with an
highest mean flow velocity (mFV) in MCA was al- ICP above 20 mm Hg (Figure 3). However, only in the
ways chosen. Normal mFV in MCA was defined as 62 ICP interval 5 to 40 mm Hg were the data normal
⫾ 12 cm/s [1]. Consequently, mFVs were consid- distributed around the regression line. Omitting re-
ered subnormal when below 50 cm/s and supernor- peated measurements, and only considering the ini-
mal when above 74 cm/s. mFVs above 120 cm/s tial TCD-ICP recording in each subject (n ⫽ 81) for
were considered severely elevated indicating vaso- the interval, 5 to 40 mm Hg R2 was 0.733 (p ⬍ 0.0001)
spasm or hyperemia [2,16]. Normal values for PI are with a SD 2.5. The regression line is ICP(PI) ⫽ 11.1
0.7 ⫾ 0.3 [20]. ⫻ PI – 1.43. Using the regression equation we get
ICP(PI) ⫽ ICP ⫾ 2.5 (SD) with a 95% confidence
CALCULATIONS AND STATISTICAL interval of ⫾4.2.
METHODS A Bland and Altman plot of all measurements
PI was calculated according to the method of Gos- confirmed that the difference between the observed
ling [6]. PI is derived from the difference in the ICP and the ICP predicted from the PI was within the
systolic and diastolic flow velocity divided by the ⫾4.2 of the observed ICP (Figure 5).
mFV. In an ICP span between 0 and 20 mm Hg, the
The PI data were correlated to ICP and CPP and method had for all measurements a sensitivity of
the correlation coefficient calculated. To control 0.88 and a specificity of 0.69, and for the first mea-
the linear correlation, the residuals were tested for surement only a sensitivity of 0.91 and a specificity
normal distribution around the regression line with of 0.79, for detecting an ICP ⬎10 mm Hg. To detect
Pulsatility Index Reflects Intracranial Pressure Surg Neurol 47
2004;62:45–51
Graph demonstrating a significant correlation between the CPP and PI with a correlation coefficient of ⫺0.493 (p
1 ⬍ 0.0001) and a correlation formula of: CPP ⫽ 89.646 – 8.258 ⫻ PI. The correlation between CPP and PI is mainly
when PI ⬎3. The dotted lines are the 95% confidence interval for the regression line, which can be significantly affected
by outliers when PI is large.
an ICP ⬎ 20 in a population with ICP between 10 and and not true volume flow, has triggered further
40, the method had for all measurements a sensi- development in the utilization of acquired data
tivity of 0.83 and a specificity of 0.99, and for the from the TCD investigations.
first measurement only a sensitivity of 0.89 and a In neurointensive care, knowledge of ICP is often
specificity of 0.92. essential. When measuring ICP the use of an intra-
ventricular catheter has for decades been consid-
ered as gold standard [11].
Discussion In the present study, we found independent of
TRANSCRANIAL DOPPLER IN intracranial pathology, a significantly strong posi-
NEUROINTENSIVE CARE tive correlation between PI and intraventricular ICP
The mobility of the equipment, the possibility of monitoring. Notably, the correlation was still strong
repeated bedside investigations together with the in the patient subgroup demonstrating mFV values
noninvasive nature of the technique, makes TCD above and below normal interval.
measurements attractive in the attempt to estimate A less clear correlation (⫺0.493) between PI and
CBF in acute care settings. The basic technical lim- CPP was seen. In head injured patients, Chan et al
itation, i.e., measuring cerebral blood flow velocity [4], reported a correlation (⫺0.725) between PI and
Graph demonstrating a significant correlation between the ICP and the PI with a correlation coefficient of 0.938 (p
2 ⬍ 0.0001) and a correlation formula of: ICP ⫽ 10.927 ⫻ PI – 1.284. The dotted lines are the 95% confidence interval
for the regression line, which can be significantly affected by outliers when PI is large.
48 Surg Neurol Bellner et al
2004;62:45–51
Graph demonstrating a significant correlation between the ICP and PI in measurements when TCD mFV exceed 120
3 cm/s. The correlation coefficient between ICP and PI was 0.828 (p ⬍ 0.002). The dotted lines are the 95% confidence
interval for the regression line, which can be significantly affected by outliers when PI is large.
CPP with an even better correlation (⫺0.942) as increases. Consequently, the TCD-PI measurement
CPP decreased below a critical value of 70 mm Hg. gives a fairly good estimation of the ICP, which
In an experimental mathematical model study, might be of value in unconscious patients.
Ursino et al [21], described relationships between
perfusion pressure, autoregulation and TCD wave- PI IN ANEURYSMAL SUBARACHNOID
form during various settings in cerebrovascular he- HEMORRHAGE
modynamics. These theoretical models indicate The correlation between ICP and PI was still intact
that during normoventilation within normal blood in SAH patients suffering vasospasm detected by
pressure range, sudden increases in PI are only TCD (Figure 3). Cerebral vasospasm in SAH is char-
attainable with a rise in ICP. Our clinical results are acterized by a high mFV through the vasospastic
in agreement with these findings. However, to rely artery, often exceeding 120 cm/s [2]. In such cases
on PI in daily practice one has to determine the there will be a decrease in CBF because of an in-
sensitivity and specificity of the method. In the crease in cerebrovascular resistance through the
acceptable ICP interval of 0 to 20 mm Hg, the TCD is artery [17]. The high flow velocity is only obtain-
highly sensitive with a medium high specificity. In able if the peripheral vascular resistance remains at
the supranormal ICP range between 10 and 40 mm a relatively low level. As a result a high mFV is
Hg the sensitivity decreases while the specificity usually not found in patients with high ICP as they
Graph demonstrating the correlation between the ICP and PI in measurements with TCD mFV below 50 cm/s. The
4 correlation coefficient between ICP and PI was 0.942 (p ⬍ 0.638). The dotted lines are the 95% confidence interval
for the regression line, which can be significantly affected by outliers when PI is large.
Pulsatility Index Reflects Intracranial Pressure Surg Neurol 49
2004;62:45–51
to the pressure created by the heart, but in the changes in cerebral perfusion pressure upon middle
arterioles the normal mean pressure drops. During cerebral artery blood flow velocity and jugular bulb
venous oxygen saturation after severe brain injury.
elevated intracranial pressure, the arterioles are
J Neurosurg 1992;77:55–61.
easily compressed creating a high peripheral vas- 5. Czosnyka M, Matta BF, Smielewski P, Kirkpatrick PJ,
cular resistance reducing the flow, and thereby, the Pickard JD. Cerebral perfusion pressure in head-
denominator of the PI. The numerator of the PI injured patients: a noninvasive assessment using
derives from the difference between systolic and transcranial Doppler ultrasonography. J Neurosurg
diastolic flow velocities. The elasticity of the normal 1998;88:802–8.
6. Gosling RG, King DH. Arterial assessment by Doppler-
vascular system dampens the flow and flow velocity
shift ultrasound. Proc R Soc Med 1974;67:447–9.
fluctuations because of blood pressure changes, 7. Guillaume J, Janny P. Manometrie intracranienne con-
which is in contrast to a rigid tube where the pres- tinue. Interet de la methode et premiers resultats.
sure is directly proportional to the flow velocity Revue Neurologique 1951;84:131–42.
[21]. A clinical setting where this is apparent is the 8. Homburg AM, Jakobsen M, Enevoldsen E. Transcra-
elevation of PI in diabetic patients with cerebral nial Doppler recordings in raised intracranial pres-
sure. Acta Neurol Scand 1993;87:488 –93.
microangiopathy and hence depressed vessel com-
9. Klingelhöfer J, Sander D, Holzraefe M. Cerebral vaso-
pliance [10]. An increased intracranial pressure al- spasm evaluated by transcranial Doppler ultrasonog-
ways results in reduced compliance of the entire raphy at different intracranial pressures. J Neurosurg
brain tissue including increased rigidity of the brain 1991;75:752–8.
arteries augmenting the velocity variations, which 10. Lee KY, Sohn YH, Baik JS, Kim GW, Kim JS. Arterial
in turn increases the denominator of the PI. Conse- pulsatility as an index of cerebral microangiopathy in
diabetes. Stroke 2000;31:1111–15.
quently, this index is sensitive to ICP changes.
11. Lundberg N. Continuous recording and control of
The flow velocity pattern in cerebral arteries is ventricular fluid pressure in neurosurgical practice.
affected both by cerebrospinal fluid pulsatility and Acta Psych Neurol Scand 1960;36(Suppl 149):1–193.
the action of cerebrovascular control mechanisms. 12. Martin NA, Patwardhan RV, Alexander MJ, et al. Char-
In a situation with severely raised ICP one might acterization of cerebral hemodynamic phases follow-
expect not only a decrease in mFV, but in particular ing severe head trauma: hypoperfusion, hyperemia,
and vasospasm. J Neurosurg 1997;87:9 –19.
an effect on velocity pattern during diastole. The 13. McQuire JC, Sutcliffe JC, Coats TJ. Early changes in
reduction in mFV during this phase is more pro- middle cerebral artery blood flow velocity after head
nounced and rapid. These factors will thus, accord- injury. J Neurosurg 1998;89:526 –32.
ing to the equation, increase the PI. This pattern is, 14. Moreno, JA, Mesalles, E, Gener, J, et al. Evaluating the
to the most extreme, seen when total brain infarc- outcome of severe head injury with transcranial
tion is present. Then mFV approaches zero and the Doppler ultrasonography. Neurosurg Focus 2000;
8(1):Article 8. http://www.neurosurgery.org/focus/
calculated PI thus reaches extreme values. jan00/8-1-8.html.
15. Nadvi SS, du Trevou MD, van Dellen JR, Gouws E. The
use of transcranial Doppler ultrasonography as a
Conclusion method of assessing intracranial pressure in hydro-
cephalic children. Br J Neurosurg 1994;8:573–7.
Our data shows a highly significant correlation be- 16. Romner B, Bellner J, Kongstad P, Sjöholm H. Elevated
tween PI and ICP independent of intracranial pa- transcranial Doppler flow velocities after severe head
thology. Accordingly, in patients with suspected injury: cerebral vasospasm or hyperemia? J Neuro-
surg 1996;85:90 –7.
increase in ICP or where an increased ICP has to be 17. Rothberg CS, Weir B, Overton TR. Treatment of sub-
excluded, PI may be of guidance and repeated PI arachnoid hemorrhage with sodium nitroprusside
measurements might prove a useful tool in neuroin- and phenylephrine: an experimental study. Neurosur-
tensive care settings. gery 1979;5:588 –95.
18. Schmidt B, Czosnyka M, Schwarze JJ, et al. Evaluation
of a method for noninvasive intracranial pressure
REFERENCES assessment during infusion studies in patients with
1. Aaslid R, Markwalder T-M, Nornes H. Noninvasive hydrocephalus. J Neurosurg 2000;92:793–800.
transcranial Doppler ultrasound recording of flow ve- 19. Steiger HJ. Carotid Doppler hemodynamics in post-
locity in basal cerebral arteries. J Neurosurg 1982;57: traumatic intracranial hypertension. Surg Neurol
769 –74. 1981;16:459 –61.
2. Aaslid R, Huber P, Nornes H. Evaluation of cerebro- 20. Steinmeier R, Laumer R, Bondár I, Priem R, Fahlbusch
vascular spasm with transcranial Doppler ultra- R. Cerebral heomdynamics in subarachnoid hemor-
sound. J Neurosurg 1984;60:37–41. rhage evaluated by transcranial Doppler sonography.
3. Bland JM, Altman DG. Statistical methods for assess- Part 2. Pulsatility indicies: Normal reference values
ing agreement between two methods of clinical mea- and characteristics in subarachnoid hemorrhage.
surement. Lancet 1986;1:307–10. Neurosurgery 1993;33:10 –9.
4. Chan KH, Miller JD, Dearden NM. The effect of 21. Ursino M, Giulioni M, Lodi CA. Relationships among
Pulsatility Index Reflects Intracranial Press Surg Neurol 51
2004;62:45–51
cerebral perfusion pressure, autoregulation, and would vary with pathology type, e.g., between sub-
transcranial Doppler waveform: a modeling study. arachnoid hemorrhage and traumatic brain injury.
J Neurosurg 1998;89:255–66.
22. Vajda Z, Buki A, Vetö F, Horvath Z, Sandor J, Doczi T. Although the correlation between pulsatility index
Transcranial Doppler-determined pulsatility index in and intracranial pressure was strong, the transcra-
the evaluation of endoscopic third ventriculostomy nial Doppler study can only be done intermittently.
(preliminary data). Acta Neurochir 1999;141:247–50. Obviously, in the pathology described, intracranial
pressure is a moving target. This limits its clinical
COMMENTARY applicability significantly. However, as the authors
Bellner et al have added to the growing body of point out, I think it could be a useful technique in
knowledge regarding transcranial Doppler sonogra- patients in whom elevated intracranial pressure
phy and its use in various neurosurgical clinical was suspected or in those patients in whom it
problems. This study has the advantage of being would be important to exclude.
prospective with a good statistical analysis of the
data. Although they are able to describe the results Gaylan L. Rockswold, M.D.
of the technique in several different pathologies, it Professor of Neurosurgery
is important to realize that the intracranial dynam- University of Minnesota
ics and the cause of increased intracranial pressure Minneapolis, Minnesota