J Neurosurg 73:699-709, 1990

Neurobehavioral outcome 1 year after severe head injury Experience of the Traumatic Coma Data Bank HARVEY S. LEVIN, PH.D., HOWARD E. GARY, JR., PH.D., HOWARD M. EISENBERG, M.D., RONALD M. RUFF, PH.D., JEFFREY T. BARTH, PH.D., JEFFREY KREUTZER, PH.D., WALTER M. HIGH, JR., PH.D., SANDRA PORTMAN, PH.D., MARY A. FOULKES, PH.D., JOHN A. JANE, M.D., ANTHONY MARMAROU, PH.D., AND LAWRENCE F. MARSHALL, M.D. Division ( f Neurosurgery, The University of Texas Medical Branch, Galveston, Texas; University of California San Diego, San Diego Head Injury Center, San Diego, California; Neuropsychology Laboratory, Department of Behavioral Medicine and Psychiatry, University of Virginia Medical Center, Charlottesville Virginia; Department of Rehabilitation Medicine, Medical College of Virginia, Richmond, Virginia; and Biometry and Field Studies Branch, National Institute of Neurological Disorders" and Stroke, Bethesda, Maryland, Department of Neurosurgery, University of Virginia Medical Center, Charlottesville, Virginia; and Department of Neurosurgery, Medical College of Virginia, Richmond. Virginia v, The outcome 1 year after they had sustained a severe head injury was investigated in patients who were admitted to the neurosurgery service at one of four centers participating in the Traumatic Coma Data Bank (TCDB). Of 300 eligible survivors, the quality of recovery 1 year after injury was assessed by at least the Glasgow Outcome Scale (GOS) in 263 patients (87%), whereas complete neuropsychological assessment was performed in 127 (42%) of the eligible survivors. The capacity of the patients to undergo neuropsychological testing 1 year after injury was a criterion of recovery as reflected by a significant relationship to neurological indices of acute injury and the GOS score at the time of hospital discharge. The neurobehavioral data at 1 year after injury, were generally comparable across the four samples of patients and characterized by impairment of memory and slowed information processing. In contrast, language and visuospatial ability recovered to within the normal range. The lowest postresuscitation Glasgow Coma Scale (GCS) score and pupillary reactivity were predictive of the 1-year GOS score and neuropsychological performance. The lowest GCS score was especially predictive of neuropsychological performance 1 year postinjury in patients who had at least one nonreactive pupil following resuscitation. Notwithstanding limitations related to the scope of the TCDB and attrition in follow-up material, the results indicate a characteristic pattern of neurobehavioral recovery from severe head injury and encourage the use of neurobehavioral outcome measurements in clinical trials to evaluate interventions for head-injured patients.

KEY WORDS head injury 9 neurobehavioral outcome Glasgow Coma Scale outcome

R

EC~Na" investigations have emphasized that neurobehavioral impairment and psychosocial maladjustment are primarily responsible for residual disability in most survivors of severe head injury, whereas focal neurological deficits such as hemiparesis are less debilitating. 7"17'22'26However, the contributions of specific neurobehavioral sequelae to overall morbidity are poorly understood. 5 Gaps in characterizing the neurobehavioral sequelae of severe head injury stem from the relatively few studies reporting longitudinal outcome data, 1-27'2~ her-

J. Neurosurg. / Volume 73 /November, 1990

9 pupillary reactivity

erogeneity in the type of injuries, 13 and inadequate analysis of cognitive performance in relation to prospectively recorded neurological indices (see Brooks, et al., 6 and Levin, et al., 18-2~for reviews). Other impediments to progress have included n o n u n i f o r m methods of characterizing the severity and type of head injury, a preexisting neuropsychiatric disorder, inadequate reporting of patients unable to complete cognitive tests or lost to attrition, and a wide variation in the measures employed to assess neurobehavioral outcome. 6 This paper characterizes the neurobehavioral out699

H. S. Levin, et aL TABLE 1

Comparison of patient demographicfeatures by completeness of 1-year outcome assessment* Demographic Feature

Completed Examinationq" Yes No

GOS Only

No Follow-Up Data

Total Cases

no. of cases 127 41 95 37 300 mean age (yrs) at injury (standard deviation) 26 (10) 30 (12) 26 (9) 26 (7) 27 (I0) education not a high-schoolgraduate 35% 22% 31% 24% 30% high-school or vocational school graduate 39% 37% 38% 38% 38.3% some college 22% 27% 20% 24% 22.3% unknown 4% 15% 12% 14% 9.3% gender female 26% 34% 16% 24% 24% male 74% 66% 84% 76% 76% race white 86% 93% 88% 89% 88% black 11% 7% 10% 8% 10% other 3% 0% 2% 3% 2% employment status (preinjury) employed (full time) 54% 61% 68% 65% 61% employed (part time) 8% 0% 4% 8% 6% student 26% 12% 14% 8% 18% unemployed 13% 20% 11% 11% 12% unknown 0% 7% 3% 11% 3% marital status married 24% 37% 28% 30% 28% not married 76% 61% 67% 65% 70% unknown 0% 2% 4% 5% 2% * Percentages are of the number of cases indicated in each column. GOS = GlasgowOutcome Scale.J6 "~Yes = completed at least one neuropsychologicaltest sufficiently to produce meaningful data; no = disabilities precluded completion.

come in survivors 1 year after treatment for severe head injury in four Traumatic C o m a Data Bank (TCDB) hospitals: the University o f Texas Medical Branch, University o f California at San Diego, Medical College o f Virginia, and University o f Virginia Medical Center. The objectives o f this paper are: 1) to define the cohort and identify limitations on generalization from the data; 2) to d o c u m e n t completeness o f the neurobehavioral examinations performed 1 year after injury; 3) to evaluate the representativeness o f the patients with complete neurobehavioral data as compared with other T C D B survivors who were eligible for this c o m p o n e n t o f the study; 4) to assess the feasibility o f administering the neuropsychological test battery to survivors o f severe traumatic coma; 5) to compare the neurobehavioral findings o f patients across the four T C D B samples; 6) to compare the neurobehavioral findings o f one sample o f T C D B survivors 1 year after injury to a matched group o f neurologically intact subjects; and 7) to investigate the relationships between neurological indices of acute injury and neurobehavioral outcome, including the patients' capacity to undergo a complete 1-year follow-up examination.

Clinical Material and Methods Patient Eligibility Entry criteria for the T C D B included a severe head injury followed within 48 hours by arrival at the emer-

700

gency r o o m of one of the four participating centers in a comatose condition (defined by a postresuscitation score o f _< 8 on the Glasgow C o m a Scale (GCS) o f Teasdale and Jennett35). This last criterion corresponded to a condition o f no eye-opening, inability to obey commands, and absence o f comprehensible speech. Exceptions to this criterion included patients who had a postresuscitation GCS score over 8 which was recorded after a delay due to early pharmacological paralysis. Cases o f delayed deterioration were also entered into the T C D B provided that c o m a ensued within 48 hours o f injury. A more detailed discussion o f these eligibility criteria is availablefl 9 Exclusion criteria for the neurobehavioral c o m p o n e n t o f the T C D B included: a history o f substance abuse which conformed to the criteria for alcohol or drug abuse set out in the Diagnostic and Statistical Manual, III; ~~ hospitalization for head injury or other cerebral insult; or psychiatric disorder. The present paper is restricted to patients who were 16 to 70 years old at the time o f injury.

Patients As shown in Table 1, 300 survivors in the T C D B were eligible for assessment o f outcome 1 year after injury. It is seen that the group is predominated by young, single white males with at least some high-school education who sustained closed head injuries in motorvehicle accidents. Table 1 shows that 85% o f the pa-

J. Neurosurg. / Volume 73/November, 1990

Neurobehavioral outcome after severe head injury tients were e m p l o y e d o r were students at the t i m e o f injury. O u t c o m e at 1 year postinjury was evaluated b y at least the Glasgow O u t c o m e Scale t6 (GOS) in 263 patients (87% o f eligible cases). A G O S score was o b t a i n e d from a structured telephone interview with a family m e m b e r for patients who d i d n o t return for follow-up e x a m i n a t i o n . O n e - y e a r o u t c o m e d a t a o f any k i n d were unavailable in 37 patients (13% o f eligible survivors) lost to follow-up review. T h e patients are differentiated in Tables 1 a n d 2 according to the c o m pleteness o f their 1-year follow-up d a t a (that is, whether the neuropsychological e x a m i n a t i o n was c o m p l e t e d or discontinued after an a t t e m p t h a d been made). Alt h o u g h the 127 patients (42% o f eligible survivors) with c o m p l e t e neuropsychological test d a t a form the basis for the b u l k o f this report, the usable o u t c o m e results for the 41 patients who were unable to c o m p l e t e the e x a m i n a t i o n due to their disabilities were also included (Table 2). A group o f 27 neurologically intact subjects was recruited t h r o u g h advertising in the local G a l v e s t o n newspaper to obtain neuropsychology d a t a for c o m p a r ison with the University o f Texas M e d i c a l Branch ( U T M B ) patients. The individual m a t c h i n g resulted in highly similar distributions o f age ( m e a n age + standard deviation: 27.9 +_ 10.0 years for patients vs. 28.9 + 11.1 years for the c o m p a r i s o n group), education (high-school graduation or less in 77% o f patients vs, 74% o f c o m parison group), a n d o c c u p a t i o n (skilled workers c o m prised 44% o f patients vs. 37% o f c o m p a r i s o n group).

Schedule of Follow- ~

Outcome Assessment

N e u r o b e h a v i o r a l o u t c o m e was assessed at baseline (that is, u p o n resolution o f p o s t t r a u m a t i c a m n e s i a or at 3 m o n t h s after injury, whichever occurred first), 6 months, 1 year, a n d 2 years following injury. T h e 1year n e u r o b e h a v i o r a l d a t a were selected for presentation in this p a p e r because this postinjury interval provides a m o r e stable level o f n e u r o b e h a v i o r a l o u t c o m e than at 6 m o n t h s which is typically characterized by rapid changes in performance. 27.2s A l t h o u g h the w i n d o w during which "1 year" o u t c o m e d a t a were accepted for entry in the T C D B ranged from 270 to 539 days, 90% o f the patients were assessed between 320 a n d 490 days. Although 2-year o u t c o m e d a t a w o u l d p r e s u m a b l y yield an even m o r e stable level o f p e r f o r m a n c e than the lyear follow-up findings,~ the size o f the s a m p l e available for analysis w o u l d be smaller due to attrition (93 rather than 127 cases),

Neuropsychological Tests Previous research identifying the m o s t salient neurobehavioral sequelae o f closed head injury has guided selection o f the o u t c o m e d o m a i n s ( m e m o r y , a t t e n t i o n / i n f o r m a t i o n processing, concept formation, n a m i n g , visuospatial ability, a n d m o t o r speed a n d c o o r d i n a t i o n ) a n d specific tests used. T h e present report is confined to measures (Table 3) selected on the basis o f previous research ~,t ~,14,15,1s-25~27to represent each d o m a i n .

J. Neurosurg. / Volume 73/November, 1990

TABLE 2

Comparison of injury-relatedfeatures by 1-year outcome completeness* Feature

Completed Examination

No GOS Only Follow-Up Total Data

Yes

No

127

41

95

37

300

56% 19% t0% 6%

60% 15% 7% 3%

60% 26% 8% 3%

60% 14% 8% 5%

58% 20 % 9% 5%

2% 5% 2%

3% 5% 7%

2% 1% 0%

5% 3% 5%

2% 3% 3%

postresuscitation GCS 3-5 6-8 9-15 untestable

20% 64% 10% 6%

68% 27% 2% 3%

19% 59% 12% 10%

19% 46% 19% 16%

26% 55% 10% 9%

pupillary reactivity both reactive one nonreactive both nonreactive unknown

76% 11% 10% 4%

39% 15% 44% 2%

81% 2% 14% 3%

78% 8% 11% 3%

73% 8% 16% 3%

13% 36%

2% 0%

15% 23%

5% 19%

11% 25%

49% 2%

24% 73%

48% 14%

49% 27%

45% 18%

259

7

10

10

25.0

26,0

25.0

9,2%

6.6%

no. of cases mechanism of injury MVA motorcycle fall/jump recreation (e.g., bicycles, skateboards) assault gunshot wound other

GOS at time of discharge good recovery moderate disability severe disability vegetative state median duration of impaired consciousnesst (day's) median 1CP during first 72 hours highest ICP (mm Hg) percent time _>20 mm Hg

7

24,0

25.0

6,0% 8.5% 6.8%

first CT scan mass:~ 15% swelling & shiftw 11% swelling, no shifiU 39% other** 27% normal 8%

22% t4% 46% 19% 0%

19% 11% 20% 33% 12%

18% 23% 38% 18% 3%

* GOS = Glasgow Outcome Scale score; MVA = motor-vehicle accident; GCS = Glasgow Coma Scale score; ICP = intracranial pressure; CT = computerized tomography. t Number of days from an initial GCS of 15 cc, a midline shift > 3 mm, or asymmetrical lateral ventricles. wDefined as compressed or absent cisterns, small ventricles, and shift > 3 mm. 1]Defined as compressed or absent cisterns, small ventricles, and no shift > 3 mm. ** Findings other than those conforming to the definition of mass and swelling. 701

H. S. Levin, et al. TABLE 3 Median neuropsychological test scores at 1 year posttrauma by TCDB Center* NeuropsychologicalTest Verbal Memory9 (no. words retrieved)

Normalt

UTMB

UCSD

MCV

UVMC

97.7(46.5)

Totalw

Center Differences[[ (P) 0.43

41.0(54.8) 53.5(80.3) 43.0(70.5) 42.0(62.8) 44.5(72.3) (n=22) (n=42) (n=20) (n=28) (n=112) Information Processing Rate (PASAT)14 0.46(0.19) 0.37(0.08) 0.34(0.11) 0.37(0.11) 0.37(0.19) 0.37(0.10) 0.25 (n=17) (n=37) (n=16) (n=18) (n=88) (no. correct/sec) Trail B Time2 (sec) 69.0(39.0) 75.0(86.0) 77.0(26.0) 89.0(84.5) 81.5(109.0) 79.0(67.0) 0.39 (n=21) (n=39) (n=29) (n=30) (n=119) Modified Card Sorting32(perseverative 1.0(2.0) 0.5(4.0) 1.0(4.5) 1.0(2.0) 1.0(2.5) 1.0(3.5) 0.64 (n=22) (n=41) (n=29) (n=29) (n=121) errors) Visual Naming4 (adjusted score) 54.0(8.0) 52.5(14.0) 55.0(13.3) 50.0(13.5) 52.0(11.3) 52.0(12.0) 0.39 (n=22) (n=42) (n=33) (n=32) (n=129) Visual Memory3 (no. of errors) 2.0(2.0) 4.5(6.8) 5.0(5.5) 4 . 0 ( 7 . 0 ) 4.0(5.3) 5.0(6.0) 0.81 (n=22) (n=41) (n=31) (n=30) (n=124) Block Design3s (age-correctedscale 10.0,(2.0) 10.0(4.0) 10.0(4.0) 8 . 5 ( 5 . 0 ) 9.0(5.8) 9.0(5.0) 0.32 (n=23) (n=41) (n=28) (n=28) (n=120) score) Grooved Pegboard~7'3~ (time in sec) 66.5,(17.8) 73.0(31.0) 74.0(39.0) 83.5(57.5) 76.0(74.0) 77.0(39.0) 0.38 (n=19) (n=39) (n=26) (n=7) (n=91) * Interquartile ranges and number of cases (n) are presented in parentheses. The interquartile range is the difference between the 25th and 75th percentile points. TCDB = Traumatic Coma Data Bank; UTMB = University of Texas Medical Branch; UCSD = University of California at San Diego; MCV = Medical College of Virginia; UVMC = University of Virginia Medical Center; PASAT = Paced Auditory Serial Addition Test. t See references for sources of normative data. * Mean was used because of unavailability of median in literature. wIncludes 127 patients with complete 1-yearexaminations and 41 patients whose disabilities precluded completion of the l-year examination. [[ Kruskal-Wallistest to determine differences among centers on neuropsychologicaltest performance.

Verbal m e m o r y was measured by two equally difficult, parallel forms~ 5 o f the Selective Reminding Test 9'~5 which were employed in alternating sequence o f examinations (that is, baseline and at 6, 12, and 24 months postinjury). Following oral presentation o f the entire 12-word list on the first trial, the examiner reminded the patient on the subsequent 11 trials o f only those words that had not been recalled on the previous trial. The n u m b e r o f words consistently retrieved from long-term m e m o r y (CLTR) across trials was analyzed. The square root o f C L T R was found to have a more normal distribution and more equal variances across groups. All subsequent analyses, then, are based on this transformed value. Visual m e m o r y was evaluated by the Benton Visual Retention Test. 3 After viewing each stimulus picture for 10 seconds, the patient was asked to draw it from m e m o r y . The n u m b e r o f errors, which has a wider range than the n u m b e r correct (0 to 10), provided a measure o f visual reproductive memory. A parallel form was used to assess copying while the designs were held in view. The Paced Auditory Serial Addition Test (PASAT) ~4 evaluated the patient's capacity to add single-digit n u m bers presented aurally at progressively shorter intern u m b e r intervals. The information-processing rate is the reciprocal o f the time (in seconds) taken by the patient for each correct response. Trail B, 2 a paper and pencil test that evaluates the speed at which the patient can draw a line connecting scattered circles while alternating between alphabetical and numerical sequences, was also used to assess speeded performance. Concept formation was evaluated by measuring per702

severative errors on the modified card sorting test. 32 This measure reflects the inability o f the patient to shift from one principle o f sorting 48 cards (such as n u m b e r or form) to a different strategy (such as matching by color). This type o f task has been shown to be differentially sensitive to frontal lobe damage. 3~'32 The Visual N a m i n g subtest o f the Multilingual Aphasia Examination a was used to evaluate the ability to name line drawings o f familiar objects on confrontation. The score is corrected for age, gender, and educational level. The Block Design subtest o f the Wechsler Adult Intelligence Scale - - Revised 38 was administered to evaluate visuospatial ability. The patient used one or both hands to assemble blocks to match the configuration o f designs presented in pictures. The Grooved Pegboard test was selected as a measure o f fine m o t o r coordination and speed. Each hand with normal strength was tested. This report is based on the data obtained using the patient's preferred hand. Overall Quality o f Outcome The five o u t c o m e categories o f the GOS were rated by the investigators according to objective criteria.16 Statistical Methods The median and interquartile range were employed as measures of central tendency and dispersion o f scores, respectively. Box plots were used to depict the distribution o f GCS scores for each category of the GOS. Logistic and linear regression was used to analyze the relationships between the neurological variables (for example, GCS score and pupillary reactivity) and at. Neurosurg. / Volume 73/November, 1990

Neurobehavioral outcome after severe head injury TABLE 4 Median neuropsychological test scores at 1 year posttrauma in Galveston patients and a comparison group of normal subjects* NeuropsychologicalTest

UTMB+

Comparison Group

Sign Test (one-tailed p)

Verbal Memory9'~6(no. wordsretrieved) 41.0(53.0)(n=22) 69.0(44.0Xn=22) < 0.01 0.37(0.07Xn= 17) 0.48(0.15)(n= 17) 0.03 Information ProcessingRate~4(no. correct/see) "Frail B Timer (see) 84.5(85.0)(n=20) 70.5(35.0)(n=20) < 0.01 Modified Card Sorting32(perseverativeerrors) 1.0(4.0Xn=21) 1.0(2.0Xn=21 ) 0.30 Visual Naming4 (adjusted score) 53.0(12.0Xn=21) 58.0(8.0)(n=21) 0,06 Visual Memory3(no. of errors) 5.0(5.0)(n=21) 2.0(4.0)(n=21) 0.02 Block Design3s (age-correctedscale score) 9.5(4.0~n=22) 9.5(4.0Xn=22) 0.32 Grooved Pegboard3~ (time in see) 73.0(31.0)(n= 19) 69.0( lO,O)(n=l 9) O.l 6 * Interquartile rangesand number of cases (n) are presented in parentheses. The interquartite range is the differencebetween the 25th and 75th percentile points. t UTMB = Universityof Texas Medical Branch. Data on visual and motor tasks were considered valid for one patient who was deaf prior to his injury. However,his data are eliminatedfrom this analysisdue to the unavailabilityof a suitable control. Fewer UTMB patientsare represented than in Table 7 because comparison subjects have not yet been recruited for patients with Katz Adjustment Scales data who were unable to complete the neuropsychologicaltesting.

outcome measures, including the GOS and neurobehavioral tests. To account for possible differences in performance across the four cohorts of head-injured patients, the treatment center was included as a covariate in the analyses. Rank tests were used for comparing neuropsychological results among groups. Results Completeness o f 1- Year O u t c o m e Data and Capacity to Undergo E x a m i n a t i o n Among the survivors, a question of potential bias is introduced by the varying completeness of outcome data 1 year after injury. Thus, we compared the demographic and clinical features of patients according to the degree of completeness of their t-year outcome examinations (see Tables 1 and 2). As seen in Table l, students were more likely to have completed their lyear examinations than were individuals who were fully employed at the time of injury'. Apart from employmerit status at injury, demographic features were unrelated to completeness of the 1-year follow-up assessment. Degree of disability (GOS) on discharge from the hospital was generally related to the completeness of the t-year outcome data (Table 2). More than twothirds of the patients who, at t year, were unable to complete the neurobehavioral assessment (that is, could not perform a portion of even a single test sufficient to produce meaningful data) were discharged from the hospital in a vegetative state as compared to 2% of patients from whom complete 1-year follow-up data were collected, Of the patients for whom 1-year outcome data were unavailable or were limited to the GOS score, 14% to 27% were vegetative at the time of their hospital discharge. Neurobehavioral Outcome Data I Year After Injury by Center The median and interquartile ranges for the neurobehavioral and psychosocial outcome measures obJ. Neurosurg. / Volume 73 / November, 1990

tained at 1 year after injury are compared for the four TCDB centers in Table 3. All of the complete data available for each test are summarized in Table 3; the first column contains the medians and interquartile ranges of scores which were previously reported for neurologically intact subjects. The values are based on samples whose demographic makeup is likely to differ from that of the TCDB patients. As summarized in the last column of Table 3, the median scores did not differ across the four samples of TCDB patients. The data of the comparison group selected to individually match each of the Galveston TCDB patients are summarized in Table 4. Verbal learning and memory (Selective Reminding Test) and reproduction of geometric designs from m e m ory (Benton Visual Retention Test) were clearly impaired in all four groups of head-injured patients compared with data collected previously for neurologically intact young adults (Table 3). The median number of words consistently retrieved by the total patient sample is 46% of the median obtained in normal subjects (Table 3). It is also seen in Table 3 that errors in visual reproductive m e m o r y by the patients exceeded the normaI level by a factor of 2.5. In agreement with the overall pattern across all four samples of patients, Table 4 shows that the TCDB patients in Galveston exhibited memory impairment on both verbal and relatively nonverbal procedures in relation to a matched comparison group. Information processing rate on the PASAT at 1 year after severe head injury was generally slowed in all four TCDB patient samples as compared with data reported for young neurologically intact adults given the same task (Table 3). This is supported further by the difference between the Galveston patients and comparison subjects (p = 0.03) (Table 4). Trail making performance was also slower in the TCDB patients than in neurologically intact subjects (Table 3), a disparity which was significant between the U T M B patients and the matched comparison group (Table 4). 703

H. S. Levin, et al. flected by block-construction skills, approximated a normal level by 1 year after injury (Tables 3 and 4). However, the variability in visuoconstructive performance was greater in the head-injured samples than in the normative data (Table 3). Speeded motor performance on a task involving placement of pegs in a grooved board tended to be slower than normal in all four TCDB patient samples (Table 3). In comparison with data collected in neurologically intact subjects, the findings 1 year after injury showed slight psychomotor slowing and increased variability of performance. Despite the tendency of more rapid pegboard test performance by the Galveston comparison group than the UTMB patients, a difference was not detected (Table 4). To summarize the pattern of 1-year outcome data presented in Table 3, memory impairment and diminished information processing speed were present in all four TCDB samples. In contrast, linguistic and visuospatial abilities had recovered relatively well by 1 year postinjury, at least in survivors capable of undergoing neuropsychological testing. This overall pattern of neuropsychological outcome is generally consistent with the comparison of the Galveston patients and matched normal subjects (Table 4) which disclosed residual disturbances in memory, information-processing speed, and speed of sequencing material (Trail B). An exception to the overall pattern across the TCDB samples is the finding of residual dysnomia in the Galveston patients when compared to a matched comparison group,

FiG. l. Box plot depicting the relationship of Glasgow Outcome Scale (GOS) score at 1 year after injury to the lowest postresuscitation Glasgow Coma Scale (GCS) score, The width of each box plot depicts the interquartile range of GCS scores for the patients whose recovery at 1 year was characterized by the corresponding GOS score. For example, the upper bound (75th percentile) of the box for good recovery is a GCS score of 7 whereas the lower bound of the box (25th percentile) is 6. The extreme scores are represented by the line extensions (for example, GCS scores of 15 and 3 for the Good Recovery group),

Extreme levels of perseveration on the modified card sorting test (as reflected by the interquartile ranges in Table 3) were more characteristic of the head-injured patients as compared to normative data collected in neurologically intact subjects of similar age (Table 3). Apart from these isolated cases of perseveration, this type of error was relatively uncommon in all four TCDB samples on this abbreviated card sorting test. As shown in Table 4, the UTMB patients and comparison group did not differ with respect to number of perseverative errors. Naming pictured objects on confrontation in each of the four TCDB cohorts approximated the level (albeit with greater variability) of neurologically intact young adults (Table 3), whereas the naming scores obtained by the patients in Galveston fell slightly below the performance of local comparison subjects (Table 4). Visuospatial ability in all four TCDB groups, as re-

Relationship Between Neurological Indices of Acute Injury and Neurobehavioral Outcome Indices of Neurological Impairment. To assess the relationship between neurobehavioral outcome 1 year after injury and early neurological indices, the GCS score and pupillary reactivity recorded by the neurosurgeon were selected because these findings were available for most patients in the TCDB and have been widely used in previous outcome studies, t'22 The following potential predictors were examined: 1) the first postresuscitation GCS score and corresponding pupillary reactivity (unavailable for 10 of the 300 eligible survivors); 2) the lowest postresuscitation GCS score and

TABLE 5 Glasgow Outcome Scalefindings by pupilla~y reactivity t year after injury* Pupillary Reactivity

Good Recovery No.

%

ModerateDisability No.

%

SevereDisability No.

%

Vegetative No.

%

first postresuscitationfindings both reactive 106 57 40 22 30 16 10 5 one or both nonreactive 14 22 13 20 27 42 10 16 worst postresuscitationfindings both reactive t01 61 35 21 24 14 6 4 one or both nonreactive 25 27 19 20 35 38 I4 15 * The data include the first postresuscitationresults and worst findings during hospitalization.The first postresuscitationpupillary reactivity was not tested in 10 of the eligiblepatients, whereasthe GlasgowOutcomeScalecategorywas unavailable for 41 patients. 704

J. Neurosurg. / Volume 73 / November, 1990

Neurobehavioral outcome after severe head injury worst pupillary reactivity; 3) intracranial pressure (ICP) during the first 72 hours; and 4) the results of the first computerized tomography (CT) scan. In determining the lowest GCS score, findings obtained while the patient was pharmacologically paralyzed were excluded. In contrast to the postresuscitation values, the worst GCS score and pupillary reactivity were selected to reflect neurological deterioration which we postulated to be more predictive of neurobehavioral recovery. Separate analyses were performed using the first postresuscitation results and the worst-case information. Outcome Measures. Outcome at 1 year after injury was assessed at three levels in relation to neurological indices of acute injury. The first was a relatively global level of functioning as measured by the GOS which was available for 259 patients. Second, we examined the ability of the patient to perform the neuropsychological test battery. Although testability was confined to the 168 patients who returned for testing, it was a more specific indicator of neurobehavioral functioning than the GOS. For the final set of analyses, we examined specific scores on the various neurobehavioral scales. These analyses were naturally limited to the 127 patients who returned for an examination and were able to complete the tests. Since few patients committed perseverative errors on the Modified Card Sorting Test, this measure was not included in further analyses. For each outcome measure, relationships with GCS score and pupillary reactivity, and the interaction of the two were examined. For these analyses, pupillary reactivity was classified as "both reactive" or "one or both nonreactive." Also, center effects (such as differences in performance) were included as a covariate for each of the analyses. The results of each set of analyses are discussed separately. Glasgow Outcome Scale. The box plot depicted in Fig. 1 shows that the lowest postresuscitation GCS score was directly related to quality of recovery at 1 year as determined by the GOS score. While patients who remained vegetative had a relatively constricted range of GCS scores, indicating deep coma (note the relatively narrow band for GCS scores depicted by the horizontal line in Fig. 1), there was wide variation in the postresuscitation level of consciousness in patients who attained a good outcome by 1 year (note extension of the horizontal line from the box plot for the "good recovery" group in Fig. 1). Analysis of the first postresuscitation GCS score (which was not necessarily the lowest GCS score) disclosed a similar relationship with outcome at 1 year. Table 5 shows that the proportions of patients with good recovery or moderate disability are smaller among patients with an abnormally reactive pupil after resuscitation while the proportions of vegetative survivors or patients with severe disability are larger. This pattern is also observed for the worst pupillary reactivity (Table 5). These patterns are supported by the results of a general logistic regression analysis (Table 6). J. Neurosurg. / Volume 73 / November, 1990

TABLE 6

Summary of logistic regression analyses for 1 year GOS* Variable

1 Year GOS x 2

df

p

1 Year Testability X2

df

first postresuscitation GCS 29.11 3 0.00001 21.29 3 pupillaryreactivity 10.32 3 0.01 7.06 3 GCS • pupillary 1.82 3 0.61 3.45 3 reactivity worst postresuscitation value GCS 37.02 3 0.00001 19.90 3 pupillaryreactivity 12.94 3 0.0048 12.27 3 GCS • pupillary 10.32 3 0.02 1.66 3 reactivity * All analyses included center as a covariate. GOS = Outcome Scale score; GCS = GlasgowComa Scale score.

p

0.0001 0.07 0.33 0.0002 0.006 0.65 Glasgow

Testability o f the Patient at 1 Year. As was seen in Table 2, patients who were unable to complete the neuropsychological test battery at 1 year were more likely to have had a nonreactive pupil and lower postresuscitation GCS scores than those patients who completed the examination (note that pupillary reactivity at the time of hospital admission was testable in 290 patients and the exact GCS score was analyzed in the logistic model rather than the scoring intervals of 3 to 5, 6 to 8, and 9 to 15). The 97 patients who had at least one nonreactive pupil at the time of hospital admission were approximately six times more likely to be unrestable 1 year after injury than the 219 patients with normal pupillary reactivity (chi-square = 17.11, p = 0.0001). A similar pattern was found for the worst postresuscitation variables (that is, indices recorded at any time after injury) using logistic regression analyses, which supported the relationship between testability and both the first postresuscitation GCS score and pupillary reactivity (Table 6). The nonsignificant interactions indicate that the effects of abnormally reactive pupils were indistinguishably different across GCS values. In particular, after adjusting for center differences, patients who had at least one nonreactive pupil at any time following postresuscitation were 3.6 times as likely to be untestable as patients with normally reactive pupils regardless of their GCS score. In contrast to the positive findings for GCS score and pupillary reactivity, Table 2 shows no impressive relationship between intracranial hypertension and testability at 1 year after injury. Logistic regression analysis of the percentage of time ICP was 20 m m Hg or greater during the first 72 hours postinjury confirmed the impression of no relationship to testability a year later after adjusting for center differences (chi-square = 0.24, p = 0.62). A similar analysis of m a x i m u m ICP was also negative (chi-square = 0.05, p = 0.83). Neurobehavioral Measures at 1 Year. The analyses of the neurobehavioral tests were similar to those of 705

H. S. Levin, et al. TABLE 7 Summary of analyses (p values) of neurobehavioral measures at 1 year after injury* NeuropsychologicalTest

First Postresuscitation

Worst Postresuscitation

GCS PR GCS • PR GCS PR GCS • PR Verbal Memory9 (CLTR) 0.38 < 0.05 0.21 0.12 0.001 0.0 l PASAT~4 0.34 0.42 0.62 0.13 0.13 0.21 Visual Naming4 0.19 0.02 0.05 0.02 0.004 0.01 Trail B Time2 0.02 0.01 0.04 0.03 0.12 0.21 Visual Memory3 (BVRT) 0.01 0.03 0.05 0.07 0.16 0.17 Block Design3s 0.07 0.003 0.01 0.04 0.03 0.04 Grooved Pegboard~7'3~ 0.24 0.18 0.66 0.06 0.62 0.49 * All analyses included center (University of Texas Medical Branch, University of California at San Diego, Medical College of Virginia, and University of Virginia Medical Center) as a covariate. BVRT = Benton Visual Retention Test; CLTR = consistent retrieval from long-term memory; GCS = Glasgow Coma Scale score; PR = pupillary reactivity.

GOS and patient testability. These outcomes were analyzed using general linear model (unweighted) procedures (Table 7). In view of the departure from a normal distribution o f the test data, it is recognized that the probability values should be regarded as estimates. Both the first and worst postresuscitation GCS scores were related to a highly speeded measure (Trail B Time), whereas only the initial GCS score was related to Visual Memory. Table 7 shows that the worst postresuscitation GCS score was significantly related to naming pictures o f objects and block construction. Pupillary reactivity recorded initially after resuscitation or at the time of the worst GCS score was strongly related to performance on several neuropsychological tests (Table 7). The relationship between the lowest postresuscitation GCS score and pupillary reactivity reflected in Table 7 is more impressive than the first postresuscitation findings. It is also seen in Table 7 that these indices o f worst neurological condition interact significantly for Verbal M e m o r y and Visual Naming, and approach significance for Visual M e m o r y and Block Design. This interaction is illustrated by the plot o f the square root o f C L T R (from the Selective Reminding Test o f Verbal Memory)

score against the lowest GCS scores presented in Fig. 2. The results o f further analysis indicate that the association between C L T R and GCS scores is significant only a m o n g patients with one or both pupils abnormally reactive. The interaction of lowest GCS score with pupillary reactivity remained significant after the patient with the highest GCS score was deleted from the analysis. Inspection of Fig. 2 left revealed no major deviation from a linear relationship between the lowest GCS score and verbal m e m o r y 1 year later. These findings indicate that very severe impairment o f consciousness concomitant with nonreactive pupils is prognostic o f neurobehavioral deficit, whereas the GCS score is far less predictive of m e m o r y 1 year later in patients with normally reactive pupils. In contrast to the predictive features o f the GCS and pupillary reactivity, type o f injury (as reflected by the CT scan obtained during the initial hospitalization) and ICP were unrelated to neurobehavioral o u t c o m e 1 year postinjury. O f the 127 patients with complete neurobehavioral examinations at 1 year, 101 had CT scans during their initial hospitalization which were available for analysis and could be classified as showing diffuse

FIG. 2. Square root of consistent long-term retrieval (CLTR) score at 1 year postinjury plotted against the lowest postresuscitation Glasgow Coma Scale (GCS) score for patients whose pupils remained normally reactive after resuscitation (left) and for those patients who had one or both pupils nonreactive following resuscitation (right). 706

J. Neurosurg. / Volume 73/November, 1990

Neurobehavioral outcome after severe head injury swelling (compressed cisterns, small symmetrical ventricles, no intracerebral lesion >__ 15 cc, and no mass effect), mass effect (shift > 3 mm), or other abnormality (scans which did not conform to the criteria for either of the other two groups). Analysis disclosed no effect which approached significance of a patient's CT group on any of the neurobehavioral tests, including verbal memory (F(2,95 df) = 0.3 i, p > 0.74), and the results for a speeded test (Trail B) were also unimpressive (F(3,104) = 0.31, p > 0.73). Intracranial pressure, which was analyzed both according to the occurrence of any recorded value 20 mm Hg or greater and the percent of time during the first 72 hours postinjury that recordings reached or exceeded this level, was also unrelated to any of the neurobehavioral outcome measures included in this paper. These negative findings are reflected by the nonsignificant effect of percent of time ICP was 20 mm Hg or greater on verbal memory (F(1,97) = 0.89, p > 0.35). The weak relationship between the percent of time that ICP was 20 mm Hg or greater and verbal memory 1 year after injury is depicted in Fig. 3. Analysis of the highest ICP recorded during the first 72 hours also revealed no effect on verbal memory at l year (F(1,94) = 0.18, p > 0.67) or on any other neurobehavioral measure. Similarly, the recording of any ICP value of 20 mm Hg or greater was unrelated to verbal memory (F(1,97) = 0.001, p > 0.96) or to any neurobehaviorat measure. Elevated ICP was common in this sample despite clinical intervention as reflected by a median of 7% time at or more than 20 m m Hg. This index of high ICP was also not significant for Trail B Time (F(1,86) = 1.59, p > 0.21).

Discussion

Limitations of the Study and Application of the Findings The patient groups from the four TCDB centers were generally comparable with respect to demographic features, with some differences in the neurological input variables. These patients, however, were not selected to be representative of all severely head-injured patients (for example, patients with antecedent neuropsychiatric disorders were excluded), so the present results should not be interpreted as typical of what might be found in other settings or under other conditions. A major problem with generalizing results is the undefined neurobehavioral outcome for those patients who did not return for evaluation 1 year postinjury (Table 1). The compliant and noncompliant patients cannot be compared on the basis of unmeasured factors, so that if bias has been introduced, the magnitude and direction of that bias are unknown. Moreover, the effects on the 1-year outcome data of various interventions received by the patients after discharge from a TCDB hospital are unknown. Since the analyses reported here were largely exploratory, a number of statistical tests were performed in examining the relationships between the outcome measures and the various predictors. As the

J. Neurosurg. / Volume 73/November, 1990

number of statistical tests increases, the likelihood of a spuriously significant finding also increases. No attempt was made here to adjust for this increase. A contribution of the TCDB is the accounting for all patients, irrespective of whether neurobehavioral measures were obtained. Our results highlight the varying completeness of the 1-year follow-up data in survivors who were eligible for the neurobehavioral component of the TCDB study. In comparison with the patients for whom follow-up data were limited to a GOS level or who were lost to follow-up review at 1 year, the survivors with complete neurobehavioral examinations had similar demographic features and neurological indices of injury, but were less frequently in a vegetative state at the time of hospital discharge.

Testability and Pattern of Neurobehavioral Recovery The capacity to undergo a 2-hour series of neuropsychological tests was a marker of recovery among patients who were examined at 1 year after injury. Severity of acute injury (as reflected by the GCS score), duration of coma, pupillary reactivity, and presence of an intracranial lesion producing mass effect were strongly related to the capacity of the patient to undergo neuropsychological examinations 1 year later. Moreover, discharge from the hospital in a vegetative condition was also a common antecedent in patients who were incapable of completing the neuropsychological tests at 1 year. Provided that patients were able to fully participate, they produced data which were considered to represent a valid estimate of their capacities. Contrary to the view that severe head injury invariably results in chronic, pervasive neuropsychological impairment, the configuration of 1-year outcome data indicates that memory deficit is especially prominent in relation to the other functions assessed in the TCDB.

FIG. 3. Square root of consistent long-term retrieval (CLTR) score at 1 year postinjury plotted against the highest intracranial pressure (ICP) recorded during the first 72 hours after injury. The negligible relationship depicted here is similar to the plots relating ICP to other neurobehavioral measures.

707

H. S. Levin, et aL Our finding that memory for new information was unequivocally impaired at 1 year after injury has important implications for quality of life in survivors of severe head trauma. Brooks and coworkers 8 recently analyzed the neurobehavioral sequelae which were most strongly related to a patient's return to work within the first 7 years after severe head injury'. Of the various neuropsychological test scores, verbal memory was found in a multiple-regression analysis to be most predictive of return to work. Consequently, the prominence of memory deficit and its strong relationship to neurological indices of acute injury may be a major factor in the chronic disability exhibited by the TCDB patients in the present study. Speeded performance, as reflected by the time required to complete Trail B ~ and serial addition under paced conditions, ~4 was also vulnerable.

Prediction of Neurobehavioral Outcome Neurological features of the early clinical course, including the lowest GCS score and pupillary reactivity, were more predictive of neurobehavioral outcome at 1 year than the first postresuscitation GCS score. This finding is in agreement with the results of Alexandre, et al., ~ who reported that the 24-hour GCS score was more predictive of neurobehavioral outcome than the GCS score obtained on admission. Our finding that the lowest GCS score predicted neurobehavioral outcome in the presence o f abnormal pupillary reactivity is compatible with a centripetal model of injury effects, in which the upper brain stem is affected immediately or as a consequence o f neurological deterioration in the most severe injuries) 3 With preservation of pupillary reactivity, a low GCS score might reflect transient changes in ICP, medications, or other phasic influences in a manner different from the unequivocal neurological deterioration associated with nonreactive pupils. This interaction of GCS score with pupillary status, which emerged for verbal memory retrieval and naming ability, has not been reported previously. The relationship between these indices of neurological deterioration, ICP, and type of injury (that is, mass lesion vs. diffuse injury) is under investigation. In contrast to the predictive features of the GCS score and pupillary reactivity, 1CP and features of the CT scan (that is, indicative of diffuse cerebral swelling vs. mass lesion) were unrelated to neurobehavioral outcome 1 year after injury in the present study. These negative findings for features of the CT scan and ICP stand in contrast to results in other studies indicating their relationship to residual neurobehavioral deficit 36'37 and to the incidence of mortality in the TCDB series, tz Uzzell and coworkers 37 reported that ICP exceeding 20 ram Hg was related to greater impairment of memory after severe closed head injury as compared to findings in survivors who had no intracranial hypertension. However, their study involved a smaller sample size, which was evaluated at varying intervals (including the initial hospitalization) during the first 6 months postin708

jury as compared with the relatively uniform 1-year follow-up period in the present study.

Implications for Multicenter Clinical Trials Notwithstanding the foregoing limitations of this study, our collection of data from testable survivors selected from consecutive admissions to four university neurosurgery services should assist extrapolation to other acute neurotrauma centers. In view of the heterogeneous pathophysiology inherent in head injury which results in small subgroups of patients (for example, those with mass lesions) at individual neurotrauma centers, the multicenter collaboration represented by the TCDB could be adopted for clinical trials to compare the initial and long-term outcome using different treatment protocols (such as techniques for controlling ICP). This strategy of outcome research could also be adopted for investigating the effects of rehabilitation at centers following a uniform treatment protocol. References

1. Alexandre A, Colombo F, Nertempi P, et al: Cognitive outcome and early indices of severity of head injury. J Neurosurg 59:751-761, 1983 2. Army Individual Test Battery: Manual of Directions and Scoring. Washington, DC: War Department, Adjutant General's Office, 1944 3. Benton AL: The Visual Retention Test: Clinical and Experimental Applications. New York: Psychological Corp, 1974 4. Benton AL, Hamsher KS: Multilingual Aphasia Examination. Iowa City: University of Iowa, 1976 (manual revised, 1978) 5. Brooks DN, Aughton ME, Bond MR, et al: Cognitive sequelae in relationship to early indices of severity of brain damage after severe blunt head injury. J Neurol Neurosurg Psychiatry 43:529-534, 1980 6. Brooks DN, Deelman BG, van Zomeran AH, et al: Problems in measuring cognitive recovery after acute brain injury. J Clin Neuropsychol 6:71-85, 1984 7. Brooks DN, McKinlay W: Personality and behavioural change after severe blunt head injury - - a relative's view. J Neurol Neurosurg Psychiatry 46:336-344, 1983 8. Brooks DN, McKinlay W, Symington C, et al: Return to work within the first seven years of severe head injury. Brain Injury 1:5-19, 1987 9. Buschke H, Fuld PA: Evaluating storage retention and retrieval in disordered memory, and learning. Neurology 24:1019-1025, 1974 10. Desk Reference to the Diagnostic Criteria from DSM-IIL Washington, DC: American Psychiatric Association, 1982 1I. Dikmen S, Reitan RM, Temkin NR: Neuropsychological recovery in head injury. Arch Neural ,t0:333-338, 1983 12. Eisenberg HM, Gary HE Jr, Aldrich EF, et al: Initial CT findings in 753 patients with severe head injury. A report from the NIH Traumatic Coma Data Bank. J Neurosurg 73:688-698, 1990 13. Gennarelli TA, Spielman GM, Langfitt TW, et al: Influence of the type of intracranial lesion on outcome from severe head injury. A multicenter study using a new classification system. J Neurosurg 56:26-32, t 982 14. Gronwall D, Wrightson P: Delayed recovery, of intelor. Neurosurg. / Volume 73/November, 1990

Neurobehavioral outcome after severe head injury

15. 16. 17.

18. 19.

20. 21. 22. 23.

24.

25. 26.

27.

28.

29.

lectual function after minor head injury. Lancet 2: 605-609, 1974 Hannay H J, Levin HS: The Selective Reminding Test: an examination of the equivalence of four forms. J Clin Exp Neuropsychol 7:251-263, 1985 Jennett B, Bond M: Assessment of outcome after severe brain damage. A practical scale. Lancet 1:480-484, 1975 Jennett B, Snoek J, Bond MR, et al: Disability after severe head injury: observations on the use of the Glasgow Outcome Scale. J Neurol Neurosurg Psychiatry 44: 285-293, 1981 Levin HS: Neurobehavioral sequelae of head injury, in Cooper PR (ed): Head Injury, ed 2. Baltimore: Williams & Wilkins, 1987, pp 442-463 Levin HS: Part II: Neurobehavioral recovery, in Becker DP, Povlishock JT (eds): Central Nervous System Trauma Status R e p o r t - 1985. Bethesda, Md: National Institutes of Health, National Institute of Neurological and Communicative Disorders and Stroke, 1985, pp 281-299 Levin HS, Benton AL, Grossman RG: Neurohehavioral Consequences of Closed Head Injury. New York: Oxford University Press, 1982 Levin HS, Grossman RG, Kelly PJ: Aphasia disorder in patients with closed head injury. J Neurol Neurosurg Psychiatry 39:1062-1070, 1976 Levin HS, Grossman RG, Rose JE, et al: Long-term neuropsychological outcome of closed head injury. J Neurosurg 50:412-422, 1979 Levin HS, High WM Jr, Goethe KE, et al: The neurobehavioral rating scale: assessment of the behavioral sequelae of head injury by the clinician. J Neurol Neurosurg Psychiatry 50:183-193, 1987 Levin HS, High WM Jr, Goldstein FC, et al: Sustained attention and information processing speed in chronic survivors of severe closed head injury. Scand J Rehab Med 17:33-40, 1988 Lezak M: Neuropsychological Assessment, ed 2. New York: Oxford University Press, 1983, pp 395-402 Livingston MG, Brooks DN, Bond MR: Three months after severe head injury: psychiatric and social impact on relatives. J Neurol Neurosurg Psychiatry 48:870-875, 1985 Mandleberg IA: Cognitive recovery after severe head injury. 3. WAIS verbal and performance IQ's as a function of post-traumatic amnesia duration and time from injury. J Neurol Neurosurg Psychiatry 39:1001-1007, 1976 Mandleberg IA, Brooks DN: Cognitive recovery after severe head injury. I. Serial testing on the Wechsler Adult Intelligence Scale. J Neurol Neurosurg Psychiatry 38: 1121-1126, 1975 Marshall LF, Becker DP, Bowers SA, et al: The National Traumatic Coma Data Bank. Part 1: Design, purpose,

J. Neurosurg. / Volume 73/November, 1990

goals, and results. 3 Neurosurg 59:276-284, 1983 30. Mathews CG, Klove H: Instruction Manual for the Adult Neuropsychology Test Battery. Madison, Wisc: University of Wisconsin Medical School, 1964 31. Milner B: Effects of different brain lesions on card sorting. Arch Neurol 9:90-100, 1963 32. Nelson HE: A modified card sorting test sensitive to frontal lobe deficits. Cortex 12:313-324, 1976 33. Ommaya AK, Gennarelli TA: Cerebral concussion and traumatic unconsciousness: correlation of experimental and clinical observations on blunt head injuries. Brain 97:633-654, 1974 34. Reitan RM, Davison LA (eds): Clinical Neuropsychology: Current Status and Applications. New York: VH Winston & Sons, 1974 35. Teasdale G, Jennett B: Assessment of coma and impaired consciousness. A practical scale. Lancet 2:81-84, 1974 36. Uzzell BP, Obrist WD, Dolinskas CA, et al: Relation of visual field defects to neuropsychological outcome after closed head injury. Aeta Neurochir 86:18-24, 1987 37. Uzzell BP, Obrist WD, Dolinskas CA, et al: Relationship of acute CBF and ICP findings to neuropsychological outcome in severe head injury. J Neurosurg 65: 630-635, 1986 38. Wechsler D: Wechsler Adult Intelligence Scale-Revised. New York: Psychological Corp, 1981 Manuscript received November 28, 1989. Accepted in final form May 11, 1990. This work was supported by the Traumatic Coma Data Bank (TCDB) under Contracts NO 1-NS-3-2339, NO1-NS-32340, NO1-NS-3-2341, NO1-NS-3-2342, NO1-NS-6-2305, and Grant NS-21889 from the National Institute of Neurological Disorders and Stroke, The TCDB Manual of Operations, which includes the TCDB data forms, is available from the National Technical Information Service (NTIS), United States Department of Commerce, 5285 Port Royal Road, Springfield, Virginia 22161 (NTIS Accession No. PB87 228060/AS). Address for Drs. Ruff and Marshall: San Diego Head Injury Center, University of California San Diego, San Diego, California. Address for Drs. Barth, Kreutzer, and Foulkes: National Institute of Neurological Disorders and Stroke, Bethesda, Maryland. Address for Dr. Jane: Department of Neurosurgery, University of Virginia Medical Center, Charlottesville, Virginia. Address for Dr. Marmarou: Department of Neurosurgery, Medical College of Virginia, Richmond, Virginia. Address reprint requests to: Harvey S. Levin, Ph.D., Division of Neurosurgery, D-73, The University of Texas Medical Branch, Galveston, Texas 77550.

709