Académique Documents
Professionnel Documents
Culture Documents
computerized tomography
A new classification of head injury based primarily on information gleaned from the initial computerized
tomography (CT) scan is described. It utilizes the status of the mesencephalic cisterns, the degree of midline
shift in millimeters, and the presence or absence of one or more surgical masses. The term "diffuse head
injury" is divided into four subgroups, defined as follows: Diffuse Injury I includes all diffuse head injuries
where there is no visible pathology; Diffuse Injury II includes all diffuse injuries in which the cisterns are
present, the midline shift is less than 5 mm, and/or there is no high- or mixed-density lesion of more than 25
cc; Diffuse Injury III includes diffuse injuries with swelling where the cisterns are compressed or absent and
the midline shift is 0 to 5 mm with no high- or mixed-density lesion of more than 25 cc; and Diffuse Injury
IV includes diffuse injuries with a midline shift of more than 5 mm and with no high- or mixed-density lesion
of more than 25 cc. There is a direct relationship between these four diagnostic categories and the mortality
rate. Patients suffering diffuse injury with no visible pathology (Diffuse Injury I) have the lowest mortality rate
(10%), while the mortality rate in patients suffering diffuse injury with a midline shift (Diffuse Injury IV) is
greater than 50%. When used in conjunction with the traditional division of intracranial hemorrhages
(extradural, subdural, or intracerebral), this categorization allows a much better assessment of the risk of
intracranial hypertension and of a fatal or nonfatal outcome. This more accurate categorization of diffuse
head injury, based primarily on the result of the initial CT scan, permits specific subsets of patients to be
targeted for specific types of therapy. Patients who would appear to be at low risk based on a clinical
examination, but who are known from the CT scan diagnosis to be at high risk, can now be identified.
FiG. 1. Computerized tomography scans showing examples of Diffuse InjuD' II: diffuse injury with limited
shift or mass effect (left); diffuse injury with multifocal hemorrhages (center); and diffuse injury marked by
deep central hemorrhage (right). The right scan shows a hemorrhage in the midbrain without effacement into
the mesencephalic cisterns.
then changed to acute subdural hematoma with a sec- Other factors are needed in addition to the CT clas-
ondary diagnosis of intraparenchymal hemorrhage. Fig- sification for prognostic purposes, as evidenced by the
ure 6 illustrates a substantial change in the CT picture outcome differences by age for a single category, Diffuse
within 12 hours, requiring a change in patient catego- Injury II (Table 3). For this diagnosis, 39% of the
rization. Thus, the dynamic nature of the patient's patients under the age of 40 years had a good or
injury can only be characterized by serial scans, but the moderate outcome versus 8% for those over 40 years;
concept that these categories are of value in predicting conversely, the older age group had over fourfold risk
the patient's course still holds. of death in comparison to the younger age group.
The CT diagnosis was a highly significant independ-
Statistical Analysis ent predictor of mortality (p = 0.0001) when age and
The association between the GCS score and the new motor score were included in the model. This three-
classification ("unknowns" omitted) was first tested by factor predictor model showed excellent fit to data (p =
the Monte Carlo approach (that is, empirical randomi- 0.86; values close to 1 indicate good fit). On the other
zation) using the Kruskal-Wallis statistic; for this, the hand, when CT diagnosis was not included, the fit was
StatXact program was employed with 10,000 random- poor (p = 0.041). Evidently, the improvement in pre-
izations. The Monte Carlo approach was indicated be- diction increased the sensitivity of the model (ability
cause of the small numbers in some of the groups in to predict deaths) by 6% (from 53% to 59%), but speci-
the classification. Logistic regression was used to show ficity was virtually unchanged. This is understandable
the improvement obtainable by the use of the new since the specificity of the two-factor prediction model
classification (excluding brain-stem and unknown in- was already very high (87%). The models themselves
juries) both above that using best postresuscitation mo- are not presented because, as has been shown elsewhere,
tor score and alone. Motor score was used instead of a truly comprehensive model takes into account addi-
the GCS score as a predictor because of missing obser- tional factors such as eye opening, hypotension, and
vations in the GCS categories. 3 The correlation between others.
motor score and GCS score in our study was 0.90, the There was a strong linkage between the ultimate
highest of the GCS subscores. Logistic regression was outcome and the degree of brain swelling and the degree
also used to develop a prediction model for Diffuse of midline shift as seen on the CT scans of patients
Injury III (swelling).
Results
TABLE 3
There is a striking direct relationship between out- Outcome of Diffuse Injury H at last evaluation by patient age*
come and initial CT scan diagnosis (p = 0.0002) (Table
2). Given the fact that the diagnostic categories in large Age (yrs)
Outcome
measure are a reflection of changes in brain volume, it (At Last _< 40 > 40 Totals
is tempting to conclude that what one is seeing is an Visit)
early indicator of the degree of intracranial hyperten- No. % No. % No. %
sion that is likely to occur in such patients. It is impor- good 15 10.0 0 0 15 8.5
tant to recognize that the degree of change in brain moderate 44 28.7 2 8.3 46 26.0
severe 63 41.1 9 37.5 72 40.7
bulk is a function of a series of complex phenomena vegetative 17 11.1 3 12.5 20 11.3
including the degree of impact injury, the presence or dead 14 9.1 10 41.7 24 13.6
absence of ischemia or ischemic hypoxia prior to CT totals 153 100 24 100 177 100
scanning, and perhaps a host of other factors not yet * Outcome classified by the Glasgow Outcome Scale. 4 For a de-
identified. scription of Diffuse Injury I1, see Table 1.
with nonsurgical lesions. As an adequate sample size ship between the CT scan appearance, mortality, and
was available in the category of Diffuse Injury III (swell- the frequency of elevated ICP in the population indi-
ing), a logistic regression equation was developed in this cates that the CT findings are strongly predictive of the
group to determine the factors that might be helpful likelihood of intracranial hypertension and that there
during the first 72 hours in sharpening the 6-month is a relationship, perhaps not completely defined, be-
outcome mortality rate prediction in such patients. As tween the degree of intracranial hypertension and the
shown in Table 4, the most powerful predictor of likelihood of death. Undoubtedly other factors, partic-
outcome in patients with Diffuse Injury II1 (swelling) ularly the actual impact injury in the severely injured
was the highest ICP. In contrast, within the other groups patient, determine the postinjury course. However, in
(not shown), age and motor scores were the most pow- patients with less substantial biomechanical injuries, it
erful predictors. Further serving to emphasize the ad- appears likely that early intervention might prevent the
verse consequences of high ICP in these patients, those development of other insults and improve both mortal-
in Diffuse Injury III group with good motor scores did ity and the overall quality of life. Certainly, the rather
relatively less well than the TCDB cohort as a whole in low mortality rate for institutions utilizing systematic
which postresuscitation GCS scores closely correlated treatment schemes for patients with GCS scores of 6,
with outcome. 7, or 8, when compared to results in hospitals with a
There is, of course, an interdependence between the much smaller experience in the care of such patients,
degree of motor dysfunction and the status of the brain suggests that therapeutic intervention does matter.
stem and the cisterns seen on CT scan. However, for This new categorization offers the possibility of early
the specific diagnostic categories used here, the CT scan identification of patients at risk as well as earlier pre-
often appeared to be a more accurate predictor of the diction of severely head-injured patients falling into
ultimate course of patients with absent or compressed broad outcome categories. While this concept needs to
cisterns than the patient's initial clinical examination if be tested in a large series of patients for whom predic-
the latter revealed a less severe injury. This was also our tions are made within 24 hours using the limited infor-
experience in the pilot phase of the TCDB study? mation that would then be available, our preliminary
It is beyond the scope of this article to deal with the experience in one center with this approach is prom-
specific question as to the reversibility of the CT scan ising.
findings based on therapeutic intervention. However, it The new classification of head injury also permits
is important to note that, within the TCDB, a substan- the early identification of patients potentially at high
tial reduction in mortality from elevated ICP has been risk from intracranial hypertension and allows the
reported in a preliminary fashion in patients with absent neurosurgeon the option of early intervention. Such a
or compressed cisterns who had less severe injuries classification also allows the identification of specific
(GCS scores of 6, 7, or 8). This suggests, at least for subsets of patients from within the overall grouping of
some patients, that early intervention for intracranial diffuse injury who have remarkably similar courses and
hypertension might play an important role in prevent- mortality rates when compared to patients with extra-
ing deterioration, a vegetative outcome, or death. axial and intra-axial mass lesions: these include patients
with diffuse injury with swelling (Diffuse Injury III) and
Discussion diffuse injury with midline shift (Diffuse Injury IV).
The CT classification introduced here appears to Diffuse Injury categories III and IV appear to be anal-
have significant application in the clinical care of the ogous in many ways to patients harboring acute, sizable
acutely head-injured patient. The very strong relation- hemorrhagic lesions and are logical groups for clinical
trials to test the efficacy of presently available as well
as new therapies. Furthermore, although the frequency
of diffuse head injury with midline shift (Diffuse Injury
TABLE 4
IV) was relatively low, its very high mortality rate in
Potential predictors of mortality ratefor patients with an head-injury centers that are experienced in the care of
intracranial diagnosis of Diffuse Injury 111 (swelling)*
such patients suggests that these patients represent a
Goodness- target group in which innovative therapies might first
Prediction
Predictor of-Fit1" be tested?
(p value) (p value)
highest ICP (first 72 hrs) < 0.001 0.038 References
pupil reactivity (postresuscitation) < 0.001 0.610
lowest 1CP (first 72 hrs) 0.023 0.930 1. Becker DP, Miller JD, Ward JD, et al: The outcome from
best motor score (postresuscitation) 0.510 0.970 severe head injury with early diagnosis and intensive
age 0.680 0.970 management, d Neurosurg 47:491-502, 1977
* ICP = intracranial pressure.For a definition of DiffuseInjuryIII, 2. Gennarelli TA, Speilman GM, Langfitt TW, et al: Influ-
see Table 1. ence of the type of intracranial lesion on outcome from
t By the Hosmer-Lemeshowtest, p value for each row for model severe head injury. J Neurosurg 56:26-32, 1982
fit including predictorsin its row and above(the higher the p value, 3. Hosmer DW, Lemeshow S: Applied Logistic Regression.
the betterthe fit). New York: John Wiley & Sons, 1989
4. Jennett B, Bond M: Assessment of outcome after severe consciousness. A practical scale. Lancet 2:81-84, 1974
brain damage. A practical scale. Lancet 1:480-484, 1975 10. Toutant SM, Klauber MR, Marshall LF, et al: Absent or
5. Lobato RD, Sarabia R, Cordobes F, etal: Posttraumatic compressed basal cisterns on first CT scan: ominous
cerebral hemispheric swelling. Analysis of 55 cases stud- predictors of outcome in severe head injury. J Nenrosurg
ied with computerized tomography. J Neurosurg 68: 61:691-694, 1984
417-423, 1988 1I. van Dongen KJ, Braakman R, Gelpke GJ: The prognostic
6. Luerssen TG, Hults K, Klauber M, et al: Improved out- value of computerized tomography in comatose head-
come as a result of recognition of absent or compressed injured patients. J Neurosurg 59:951-957, 1983
cisterns on initial CT scans, in Hoff JT, Betz AL (eds):
Intracranial Pressure VII. Berlin: Springer-Verlag, 1989,
pp 598-602
7. Marshall LF, Becker DP, Bowers SA, et al: The National This work was additionally supported by National Insti-
Traumatic Coma Data Bank. Part 1: Design, purpose, tute of Neurological Disorders and Stroke Contracts NO1-NS-
goals, and results. J Neurosurg 59:276-284, 1983 9-2306, 2307, 2308, 2309, and 2313 for the Pilot Traumatic
8. Teasdale E, Cardoso E, Galbraith S, el al: CT scan in Coma Data Bank.
severe diffuse head injury: physiological and clinical cor- Address reprint requests to: Lawrence F. Marshall, M.D.,
relations. J Neurol Neurosurg Psychiatry 47:600-603, Division of Neurosurgery, University of California Medical
1984 Center, 225 Dickinson Street H-501, San Diego, California
9. Teasdale G, Jennett B: Assessment of coma and impaired 92103-1990.