Determination of threshold levels of cerebral perfusion pressure and intracranial pressure in severe head injury by using receiver-operating characteristic curves: an observational study in 291 patients
Ir. Chambers et al., Determination of threshold levels of cerebral perfusion pressure and intracranial pressure in severe head injury by using receiver-operating characteristic curves: an observational study in 291 patients, J NEUROSURG, 94(3), 2001, pp. 412-416
Object. Intracranial pressure (ICP) and cerebral perfusion pressure (CPP) a
re frequently monitored in severely head injured patients. To establish whi
ch one (ICP or CPP) is more predictive of outcome and to examine whether th
ere are significant threshold levels in the determination of outcome, recei
ver-operating characteristic (ROC) curves were used to analyze data in a la
rge series of head-injured patients.
Methods. Data were obtained from a total of 291 severely head injured patie
nts (207 adults and 84 children). Outcome was categorized as either indepen
dent (good recovery or moderate disability) or poor (severely disabled, veg
etative, or dead) by using the Glasgow Outcome Scale; patients were also gr
ouped according to the Marshall computerized tomography scan classification
.
Conclusions. The maximum value of a 2-minute rolling average of ICP reading
s (defined as ICPmax) and the minimum value of the CPP readings (CPPmin) we
re then used to calculate the sensitivity and specificity of the ROC curves
over a range of values. Using ROC curves, a threshold Value for CPPmin of
55 mm Hg and for ICPmax of 35 mm Hg appear to be the best predictors in adu
lts. For children the levels appear to be 43 to 45 mm Hg for CPPmin and 35
mm Hg for ICPmax. Higher levels of CPPmin seem important in adults with mas
s lesions. These CPP thresholds (45 mm Hg for children and 55 mm Hg for adu
lts) are lower than previously predicted and may be clinically important, e
specially in children, in whom a lower blood pressure level is normal. Also
, CPP management at higher levels may be more important in adults with mass
lesions. A larger observational series would improve the accuracy of these
predictions.