- Email: [email protected]
Somatosensory evoked potentials for prediction of outcome in acute severe brain injury
Somatosensory evoked potentials for prediction of outcome in acute severe brain injury J. B e c a , MBChB, FRACP, P. N. COX, MBChB, FFARCS,FRCPC, M. J. Taylor, PhD, D. Bohn, MBBCh, FRCPC, W. Butt, MBBS,FRACP, W. J. L o g a n , MD, FRCPC, J. T. Rutka, MD, FRCSC, a n d G. Barker, MBBS,FFARACS From the Departments of Critical Care, Neurology, and Neurosurgery, Hospital for Sick Children, Toronto, Canada, and the Department of Critical Care, Royal Children's Hospital, Melbourne, Victoria, Australia The purpose of this study was to e v a l u a t e prospectively short-latency somatosensory e v o k e d potentials (SEPs) as a predictor of o u t c o m e in acute, severe brain injury, and to c o m p a r e this with the predictive power of the motor c o m p o nent of the Glasgow Coma Scale score and c o m p u t e d t o m o g r a p h i c scan. Outc o m e was measured with the Glasgow O u t c o m e Scale at a minimum of 6 months after injury. We studied 109 patients ( a g e d 0.1 to 16.8 years) with SEPswithin 4 days of the onset of coma. Four patients had absent SEPs and a favorable o u t c o m e by the Glasgow O u t c o m e Scale (full recovery or m o d e r a t e disability); two of these patients had meningitis with bilateral subdural effusions, one had a midbrain hemorrhage, and one had a decompressive c r a n i e c t o m y for uncontrolled intracranial hypertension. Normal SEPs had a positive predictive value for favorable o u t c o m e of 93% (95% c o n f i d e n c e interval (Cl), 77% to 99%), and absent SEPs had a positive predictive value for unfavorable o u t c o m e by the Glasg o w O u t c o m e Scale (severe disability, survival in a persistent v e g e t a t i v e state, or death) of 92% (95% Cl, 80% to 98%). If the a b o v e identifiable clinical situations in which a physical barrier existed to i m p e d e cutaneous reception of the electrical impulse were e x c l u d e d , the positive predictive value of absent SEPs for poor o u t c o m e r e a c h e d 100% (95% Cl, 92% to 100%). An absent motor response to painful stimulus also had 100% positive predictive value (95% Cl, 84% to 100%) for unfavorable outcome; however, 23% of patients could not be e v a l u a t e d because of the effects of muscle relaxants or sedatives. In patients with traumatic brain injury, results of c o m p u t e d t o m o g r a p h y did not reliably predict outcome. Of the 59 patients with unfavorable outcome, 76% could be identified with SEPs c o m p a r e d with 36% with examination of motor function. We suggest that SEPsbe performed in children with a c u t e severe brain injury b e c a u s e they a d d an important tool to the physician's prognostic armamentarium. We c o n c l u d e that in the a b s e n c e of the a b o v e mentioned identifiable clinical situations, absent SEPs predict 100% unfavorable outcome, and this finding may warrant consideration of withdrawal of treatment in children with brain injuries. (J PEDiATR1995;126: 44-9)
Submitted for publication May 9, 1994; accepted July 28, 1994. Reprint requests: P. N. Cox, MBChB, FFARCS, FRCPC, Department of Critical Care, Hospital for Sick Children, 555 University Ave., Toronto, Ontario M5G 1XS, Canada. Copyright © 1995 by Mosby-Year Book, Inc. 0022-3476/95/$3.00 + 0 9/20/59569 44
Reliable and early prediction of neurologic outcome in patients with acute brain injury remains a challenge for the intensivist; early identification of patients with a poor neurologic prognosis would enable physicians to counsel families to agree to stop futile and costly treatment. Short-latency somatosensory evoked potentials have been shown to
The Journal o f Pediatrics Volume 126, Number 1
I
CI CT SEPs
Confidence interval Computed tomographic Somatosensory evoked potentials
Beca et al.
I
provide accurate prognostic information in both adults and children with an acute brain injury. 1-8 In particular, bilateral absence of SEPs appears to be uniformly associated with a poor outcome. SEPs can be obtained relatively easily; they are noninvasive and require 30 to 40 minutes to perform. Short-latency SEPs reflect the subcortical and primary cortical aspects of the somatosensory pathways. They are not affected by the sedative and analgesic drugs used in intensive care 9-1~ and can be used in patients receiving muscle relaxants, in whom clinical assessment is impossible. Studies have been done on auditory and visual evoked potentials as predictors of outcome, but these indicators do not add further prognostic information; both may be unaffected by brain injury so that, although loss of either is associated with poor outcome, neither is reported to be as sensitive as SEPs. 1, 2, 7, 12 It has been proposed that evoked potentials be used as an adjunct in planning therapy13; Judson et al. 5 have incorporated SEPs into the management of head injury in adults. There are only a few studies of SEPs in children with brain injuries,2, 6, 7, 14 and no studies, of either pediatric or adult patients, provide details of confidence limits. The aims of this study were to evaluate SEPs prospectively as a predictor of neurologic outcome in a large number of children with acute, severe brain injury. Our hypothesis was that no child would have a good recovery if SEPs were absent bilaterally at any time. If this generalization was not true, we wished to determine whether exceptions could be defined--in particular, whether focal intracranial abnormalities could lead to loss of SEPs but the child could still make a good recovery. We also compared the predictive power of SEPs with clinical assessment and, in patients with traumatic brain injury, with computed tomographic scan. METHODS Patients. Children eligible for this study were those older than 1 month of age with severe brain injury who were admitted to the intensive care unit at the Hospital for Sick Children (HSC), Toronto, Canada, from July 1990 to February 1992, and at the Royal Children's Hospital (RCH), Melbourne, Australia, from February 1991 to February 1992. Severe brain injury was defined as (1) a Glasgow Coma Scale score <8, (2) the inability to localize painful stimuli if a full Glasgow Coma Scale was not performed (i.e., motor score <4), or (3) a history suggestive of severe brain injury in a patient who could not be examined neurologically. All patients had endotracheal tubes in place and were supported by mechanical ventilation. Intracranial
45
pressure was monitored in children with traumatic brain injury. Patients were excluded if they were known to have neurological deficits before the acute event, or if they subsequently died from nonneurologic causes, or if a diagnosis of brain death was made before SEPs were performed. We performed SEPs in 120 patients with acute traumatic and nontraumatic brain injury. Of these, 11 were excluded from further analysis either because they died from nonneurologic causes (n = 8) or had neurodevelopmental disabilities before the injury (n --- 3). The median age of the remaining 109 patients was 4.8 years (range, 2 months to 16.8 yrs): Infants younger then 1 month of age were not included because of the requirement of some history of normal neurologic development before injury (for our criteria), and we did not wish to include the range of possible perinatal injuries. The application of evoked potentials in both term and preterm infants warrants separate studies. Somatosensory evoked potentials can be reliably recorded in all normal infants if measurements used for older children are modified for those younger than 4 months 15, 16; if these modifications are not made, this is not the case. 17, 18 In our study there were only two children younger than 4 months of age (both were 2 months of age). The mechanisms of brain injury were trauma in 53 (48.6%) patients, hypoxic ischemic encephalopathy in 33 (30.3%), and other in 23 (21.1%). The latter group included meningitis (n = 6), hemorrhage (n = 3), tumor (n = 3), cardiopulmonary bypass (n = 3), hemolytic uremic syndrome (n = 2), insulin overdose (n = 1), hepatic encephalopathy (n = 1), and encephalopathy of unknown cause (n = 4). Children with acute brain dysfunction after cardiopulmonary bypass were classified as "other" because of the several possible mechanisms of injury. The Glasgow Coma Scale score or the motor score alone was recorded at the time of admission. If the patient was unable to be examined clinically because of the effects of muscle relaxants, the most recent assessment before admission was used (for example, by the referring hospital or an ambulance attendant). All patients had CT scans as part of their diagnostic evaluation. The first CT scan after admission was classified according to the Traumatic Coma Data Bank scale for patients with traumatic brain injury. 19 The outcome was assessed with the Glasgow Outcome Scale 2° at a minimum of 6 months after injury. This scale consists of five categories of patient outcomes: good (normal or mild disability), moderate (disabled but independent), severe (conscious but severely disabled and dependent), persistent vegetative state, and death. For analysis these groups were further simplified into favorable outcome (including good or moderate disability) and unfavorable outcome (including severe disability, vegetative state, or death). Outcome assessment was performed by pediatric neurologists with no knowledge of the results of the SEPs.
46
Beca etal.
The Journal of Pediatrics January 1995
Table I. Result and outcome of SEPs
and the worst SEPs recorded for each child. Confidence intervals were derived with the binomial distribution.23
Outcome
Initial SEPs Normal Unilateral delay Bilateral delay Unilateral absent Bilateral absent Worst SEPs Normal Unilateral delay Bilateral delay Unilateral absent Bilateral absent
Favorable* (No.)
Unfavorablet (No.)
31 4 5 9 1
3 0 5 10 41
26 5 4 11 4
2 0 4 8 45
*Normal or moderate disability. tSevere disability,vegetativestate, or death.
Somatosensory evoked potentials. The SEPs were recorded with a model 2000 measuring unit (Nicolet Instruments, Madison, Wis.) (HSC) or a Medilec Maestro measuring unit (Medilec, Old Woking, Surrey, England) (RCH). The right and left median nerves were stimulated at the wrist, and SEPs were recorded over the cervical spine (C7) and contralateral somatosensory cortex (C3' and Ca') with a frontal reference electrode (Fpz). If the wrist was inaccessible, stimulation was performed at the antecubital fossa. If the cervical spine was inaccessible, peripheral SEPs were recorded at the Erb point (over the brachial plexus). At least two averages of 256 artifact-free responses were recorded per arm. The bandpass was 30 to 3000 Hz (HSC) and 20 to 2000 Hz (RCH), the sweep interval was 50 ms, and the stimulation rate was 4.1 Hz. In infants younger than 4 months of age, 64 trials were averaged; the bandpass was 30 to 3000 Hz and 5 to 1500 Hz (HSC), or 20 to 2000 Hz and 3 to 1000 Hz (RCH), with a sweep of 200 ms, and a stimulation rate of 1.1 Hz. 19 The SEPs were classified into five categories: (1) normal, (2) unilateral increased latency, (3) bilateral increased latency, (4) unilateral absence, and (5) bilateral absence. Latency was considered prolonged if it was greater than 2.5 SD above normal for the child's age.iS, z1The first three categories were further subdivided into normal and low amplitude; category 4 was subdivided according to whether the unilateral response was normal or delayed. If there was uncertainty about classification, SEPs were recorded as the better alternative. When possible, SEPs were recorded within 48 hours of admission, and in all cases they were recorded within 4 days of the onset of coma. They were repeated within the first week in 67 of the 109 patients. The positive and negative predictive values, and the sensitivity and specificityz2 were calculated for the first SEPs
RESULTS For 80% of the patients, initial SEPs were performed within 48 hours, and for 91% within 72 hours; 54% of patients had SEPs repeated one to three times during the first week. The overall outcome was good in 34 (31.2%) of the 109 children, and there was moderate disability in 16 (14.7%), severe disability in 13 (11.9%), persistent vegetative state in 9 (8.3%), and death in 37 (33.9%), yielding an overall favorable outcome of 44.9%. Reduced SEP amplitude did not appear to be predictive of outcome, but numbers within these subgroups were too small for meaningful analysis. Favorable outcome was associated with normal (category 1) SEPs at initial testing in 31 of 34 children (91% positive predictive power) (Table I). If the worst SEP recording was considered, the positive predictive power increased to 93% (95% CI, 77% to 99%). Of 42 children with bilaterally absent SEPs (category 5) at initial SEP testing, 41 had an unfavorable outcome (98% positive predictive power). This included a single child in whom SEPs were bilaterally absent for anatomic reasons (subdural effusions). In 49 children, SEPs were absent in the worst recording. Of these, 45 had an unfavorable outcome. (92% positive predictive power; 95% CI, 80% to 98%); 4 of these 45 children had an evoked response (categories 1 to 4) present on the initial recording that became absent on later testing. The sensitivity, specificity, and predictive value of later versus initial SEPs were compared; minimal differences were found (Table II). An additional four patients had bilaterally absent SEPs (category 5) at either initial or after additional testing and a favorable outcome (normal or moderate disability) (Table I). Two of these four children had meningitis with subdural effusions; one had a favorable outcome and one had moderate disability at follow-up. Of the remaining two children, one had traumatic brain injury with hemorrhage in the tegmentum of the mesencephalon and the right cerebral peduncle and contusion to the posterior limb of the left internal capsule; she had a favorable outcome with moderate disability. The fourth patient also had traumatic brain injury and grade I SEPs on the first recording; the SEPs became absent after decompressive craniectomy for uncontrolled intracranial hypertension, and then returned by day 7; she also had moderate disability. Each of these four children had identifiable clinical features known to interfere technically with the recording of SEPs. If these four patients are excluded, the positive predictive value for absent SEPs is 100% (95% CI, 92% to 100%). Of the 45 children with bilaterally absent SEPs (category 5) on either initial or subsequent testing and poor outcome, 27% were alive at 6 months, but were either
The Journal o f Pediatrics Volume 126, Number 1
Beca et al.
47
Table II. Predictive value, sensitivity, and specificity for SEPs and motor score
Worst SEPs Absent (category 5)/unfavorable Normal (category 1)/favorable Initial SEPs Absent (category 5)/unfavorable Normal (category l)/favorable Motor score Flaccid/unfavorable
Positive predictive value
Negative predictive value
Sensitivity
Specificity
45/49 (92) 26/28 (93)
46/60 (77) 57/81 (70)
45/59 (76) 26/50 (52)
46/50 (92) 57/59 (97)
41/42 (98) 31/34 (91)
49/67 (73) 56/75 (75)
41/59 (69) 31/50 (62)
49/50 (98) 56/59 (95)
21/21 (100) 38/63 (60) 21/46 (46) Data are presentedas numberwith test result/numberwith outcome,Numbersin parenthesesindicatepercentages. severely disabled or in a persistent vegetative state. Of the 32 patients with abnormal but present SEPs (categories 2 to 4), 20 (63%) had favorable outcomes. All the patients with no response to pain had an unfavorable outcome (Table III). Of the 32 patients with flexor or extensor posturing responses, 14 (44%) had a favorable outcome, and 24 (77%) of 31 with withdrawal or a localizing response had a favorable outcome. The motor score could not be recorded in 25 (22.9%) of the patients, usually because muscle relaxants or large doses of sedative drugs were used, but some motor assessments were either incorrectly performed or not recorded by the admitting physicians. An absent motor response to painful stimulus had a positive predictive power of 100% (95% CI, 84% to 100%) for unfavorable outcome. However, sensitivity was 46% in those who could be clinically assessed and only 36% in all patients, compared with sensitivity of 69% for initial SEPs. No specific CT scan appearance in patients with traumatic brain injury was invariably associated with either a good or bad outcome (Table IV); 38 (72%) of the 53 patients had diffuse injury II or III (Traumatic Coma Data Bank scale), and in these groups CT scan had poor predictive power. Numbers in other categories were very small. DISCUSSION This study shows that bilaterally absent SEPs accurately predict unfavorable outcome in the setting of both traumatic and nontraumatic coma. In any test of neurologic function that attempts to predict neurologic outcome, the positive predictive power and the specificity are of much greater importance than the sensitivity. It would be highly undesirable for treatment to be withdrawn because a test incorrectly predicted a poor outcome, but less unacceptable if treatment were continued because a test incorrectly predicted a good outcome. Knowledge of the CIs is essential, to proper test use, but in published studies on SEPs these have not been provided. For a test to be clinically useful, it is necessary to know the outcome for a large number of patients with thepoor prognostic sign.24, 25The clinician needs
38/38 (100)
Table III. Motor score related to outcome: Number of children (percent) Outcome Favorable
Motor score Localizes Withdrawal Flexor Extensor Flaccid Inaccessible
Unfavorable
No.
%
No.
%
6 18 7 7 0 12
75 78 41 47 0 48
2 5 10 8 21 13
25 22 59 53 100 52
Table IV. Computed tomographic scan related to outcome of traumatic brain injury Outcome Favorable
Diffuse injury I (normal) II (minor abnormality) III (edema) IV (shift) Evacuated mass Nonevacuated mass
Unfavorable
No.
%
No.
%
1 9 12 1 5 1
50 60 52 25 83 33
I 6 11 3 1 2
50 40 48 75 17 67
to know what proportion of children with absent SEPs will have a poor outcome. In this study initial SEPs were only slightly different than additional SEPs performed later. Although many children had SEPs that showed some improvement or deterioration in latency and amplitude, only a few children had additional SEPs that crossed the present/absent boundaries. A larger series is needed to determine the significanceof the changes, such as from abnormal to normal latencies. Normal SEPs, regardless of when recorded, predicted good outcome in 91% to 93% of children; abnormal but present responses
48
Beca et al.
predicted good outcome in 63%. Absent SEPs predicted unfavorable outcome in 92% of children. Absent SEPs on later studies in patients with no identifiable anatomic cause for their absence always predicted poor outcome (95% CI, 92% to 100%). If the lower limit of the 95% CI is used, this study shows that up to 8% of children with absent SEPs could still have a favorable outcome. However, given the very high predictive power shown in many adult and a few pediatric studies, 13 the risk must be much lower than this. With Bayesian methods, the actual risk is likely to be 2% for this sample sizeY Huge numbers would be required to reduce the lower limit of the 95% CI to less than 1%, and ultimately the decision about how to use this information will depend on the individual doctor, hospital, and country. Although we have shown that absent SEPs have high sensitivity and specificity, in certain instances they cannot be relied on to predict outcome accurately; these include extraaxial collections, brain stem hemorrhage, and after cranial decompression. Extraaxial collections, such as subdural effusions in meningitis, increase the physical distance between the brain and recording electrodes, which can be sufficient to prevent the recording of responses. However, because CT scanning is part of the management of such conditions, the presence of these collections will be known at the time. Likewise, CT scanning will reveal brain stem hemorrhage, which can interfere with conduction in the subcortical somatosensory pathways. Deep cerebral hemorrhages in traumatic brain injury, especially if in the brain stem, are in themselves associated with poor outcome. Midbrain hemorrhage in adults has been shown to lead to loss of SEPs and has been uniformly associated with poor outcome.26, 27 However, because the child with brain stem hemorrhage in this study survived with moderate disability, we suggest that absent SEPs should not be used to predict outcome in cases with brain stem hemorrhage. The final exception, of cranial decompression for intracranial hypertension, has been previously reported in children with Reye syndrome.2 Taylor and Farrell7 also identified a child with brain injury as a result of a lightning strike who had transient loss of SEPs and good outcome. Both these clinical situations are encountered so rarely that it will be difficult to get further information, but in the former, SEPs recorded before decompression accurately predicted the outcome. If these conditions are considered, we have shown that absent SEPs are relatively specific in predicting poor outcome. It is essential that technically satisfactory SEPs be obtained, and recordings in which the components are obscured by artifact should not be used clinically. We have found that electrical interference is rarely a problem, even in an intensive care setting. The most common source of artifact is from muscle activity in nonparalyzed patients. Clear cervical or brachial plexus potentials, demonstrating
The Journal of Pediatrics January 1995
the integrity of the peripheral nervous system, must be obtained before absent cortical potentials are regarded as significant. Absence of motor response to painful stimuli was as specific as bilaterally absent SEPs for predicting unfavorable outcome but considerably less sensitive. In evaluating clinical assessment in this study, motor response was used for two reasons. First, it is likely that most, if not all, the predictive power of the Glasgow Coma Scale is derived from the motor score. 28 Moreover, patients with nontraumatic coma commonly had only the motor score rather than the fuli Glasgow Coma Scale score recorded. Muscle relaxants and sedative and analgesic agents are frequently used in neurointensive care; thus it is often difficult to obtain additional clinical assessments of patients with acute brain injury, but SEPs can be used in these patients. No CT scan appe~trance had absolute specificity, although assessment was limited by the small number of patients in the several subgroups. The distinction between moderate and severe disability at clinical follow-up is difficult in children, and becomes more difficult with younger age; this distinction is critical because we used it to separate favorable from unfavorable outcome in this study. However, all our patients with severe disabilities had profound developmental delay (functioning at less than half the appropriate level for age) and severe motor disabilities (preventing independent mobility). We suggest that SEPs be performed in all children with acute, severe brain injury, because they are the best predictor of outcome in this group. The SEPs should be performed within 24 to 48 hours of admission and be repeated within 24 to 48 hours to increase positive predictive power. If a CT scan excludes th6 presence of extraaxial fluid collections or brain stem hemorrhage, the presence of normal SEPs suggests a 93% chance of a favorable outcome, whereas the absence of SEPs predicts a 92% chance of an unfavorable outcome. In patients who had no physical barrier impeding the recording of the cortical response, absent SEPs predicted a 100% chance of unfavorable outcome. If SEPs are absent and clinical assessment is compatible with an unfavorable outcome, we believe that consideration should be given to withdrawal of treatment. We thank L. J. MacMillan, N. K. Keenan, and B. G. Carter for their excellent technical support. REFERENCES
1. De Meirleir DJ, Taylor MJ. Evoked potentials in comatose children: auditory brainstem responses. Pediatr Neurol 1986; 2:31-4. 2. De Meirleir LJ, Taylor MJ. Prognostic utility of SEPs in comatose children. Pediatr Neurol 1987;3:78-82. 3. Greenberg RP, Newlon PG, Becker DP. The somatosensory
The Journal o f Pediatrics Volume 126, Number 1
4. 5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
Beca et al.
evoked potential in patients with severe head injury: outcome prediction and monitoring of brain function. Ann N Y Acad Sci 1982;388:683-8. Hume AL, Cant BR. Central somatosensory conduction after head injury. Ann Neurol 1981;10:411-9. Judson JA, Cant BR, Shaw NA. Early prediction of outcome from cerebral trauma by somatosensory evoked potentials. Crit Care Med 1990;18:363-8. Lutschg J, Pfenninger J, Ludin HP, Vassella F. Brain-stem auditory evoked potentials and early somatosensory evoked potentials in neurointensively treated comatose children. Am J Dis Child 1983;137:421-6. Taylor M J, Farrell EJ. Comparison of the prognostic utility of VEPs and SEPs in comatose children. Pediatr Neurol 1989; 5:145-50. Taylor M J, Murphy W J, Whyte HE. Prognostic reliability of somatosensory and visual evoked potentials of asphyxiated term infants. Dev Med Child Neurol 1992;34:507-15. Ganes T, Lundar T. The effect of thiopentone on somatosensory evoked responses and EEGs in comatose patients. J Neurol Neurosurg Psychiatry 1983;46:509-14. Newlon PG, Greenberg RP, Hyatt MS, Enas GG, Becket DP. The dynamics of neuronal dysfunction and recovery following severe head injury assessed with serial multimodality evoked potentials. J Neurosurg 1982;57:168-77. Sutton LN, Frewen T, Marsh R, Jaggi J, Bruce DA. The effects of deep barbiturate coma on multimodality evoked potentials. J Neurosurg 1982;57:178-85. Cant BR, Hume AL, Judson JA, Shaw NA. The assessment of severe head injury by short-latency somatosensory and brain-stem auditory evoked potentials. Electroencephalogr Clin Neurophysiol 1986;65:188-95. Goodwin SR, Friedman WA, Bellefteur M. Is it time to use evoked potentials to predict outcome in comatose children and adults? Crit Care Med 1991;19:518-24. Frank LM, Furgiuele TL, Etheridge JE Jr. Prediction of chronic vegetative state in children using evoked potentials. Neurology 1985;35:931-4.
49
15. George SR, Taylor MJ. Somatosensory evoked potentials in neonates and infants: developmental and normative data. Electroencephalogr Clin Neurophysiol 1991 ;80:94-102. 16. Bongers-Schokking C J, Colon E J, Hoogland RA, Van den Brande JL, De Groot CJ. The somatosensory evoked potentials of normal infants: influence of filter bandpass, arousal state and number of stimuli. Brain Dev 1989; 11:33-9. 17. Willis J, Seales D, Frazier B. Short latency somatosensory evoked potentials in infants. Electroencephalogr Clin Neurophysiol 1984;59:366-73. 18. Laureau E, Majnemer A, Rosenblatt B, Riley P. A longitudinal study of short latency somatosensory evoked responses in healthy newborns and infants. Electroencephalogr Clin Neurophysiol 1988;71:100-8. 19. Marshall LF, Marshall SB, Klauber MR, et al. A new classification of head injury based on computerised tomography. J Neurosurg 1991;75:S14-20. 20. Jennett B, Bond M. Assessment of outcome after severe brain damage. Lancet 1975;1:480-4. 21. Taylor M J, Fagan ER. SEPs to median nerve stimulation: normative data for paediatrics. Electroencephalogr Clin Neurophysiol 1988;71:323-30. 22. Altman DG. Practical statistics for medical research. 1st ed. London: Chapman and Hall, 1990:409-16. 23. Lentner C. Geigy scientific tables, basel: Ciba Geigy, 1982:89102. 24. Bates D. Defining prognosis in medical coma. J Neurol Neurosurg Psychiatry 1991;54:569-71. 25. Shewmon DA, De Giorgio CM. Early prognosis in anoxic coma: reliability and rationale. Neurol Clin 1989;7:823-43. 26. Rosenblum WI, Greenberg RP, Seelig JM, Becker DP. Midbrain lesions: frequent and significant prognostic feature in closed heart injurt. Neurosurgery 1981;9:613-20. 27. Marshall LF, Gautille T, Klauber MR, et al. The outcome of severe closed head injury. J Neurosurg 1991;75:$28-36. 28. Jagger J, Jane JA, Rimel R. The Glasgow Coma Scale: To sum or not to sum? Lancet 1983;2:97.
AVAILABILITY OF JOURNAL BACK ISSUES As a service to our subscribers, copies of b a c k issues of THE JOURNAL OF PEDIATRICS for the preceding 5 years are m a i n t a i n e d and are available for purchase from the publisher, M o s b y - Y e a r Book, Inc., at a cost of $10.00 per issue. T h e following quantity discounts are available: 25% off on quantities of 12 to 23, and one third off on quantities of 24 or more. Please write to M o s b y Year Book, Inc., Subscription Services, 11830 Westline Industrial Drive, St. Louis M O 631463318, or call (800)453-4351 or (314)453-4351 for information on availability of particular issues. If back issues are unavailable from the publisher, photocopies of complete issues are available from U M I , 300 N. Z e e b Rd., A n n Arbor, M I 48106, (313)761-4700.
Submitted for publication May 9, 1994; accepted July 28, 1994. Reprint requests: P. N. Cox, MBChB, FFARCS, FRCPC, Department of Critical Care, Hospital for Sick Children, 555 University Ave., Toronto, Ontario M5G 1XS, Canada. Copyright © 1995 by Mosby-Year Book, Inc. 0022-3476/95/$3.00 + 0 9/20/59569 44
Reliable and early prediction of neurologic outcome in patients with acute brain injury remains a challenge for the intensivist; early identification of patients with a poor neurologic prognosis would enable physicians to counsel families to agree to stop futile and costly treatment. Short-latency somatosensory evoked potentials have been shown to
The Journal o f Pediatrics Volume 126, Number 1
I
CI CT SEPs
Confidence interval Computed tomographic Somatosensory evoked potentials
Beca et al.
I
provide accurate prognostic information in both adults and children with an acute brain injury. 1-8 In particular, bilateral absence of SEPs appears to be uniformly associated with a poor outcome. SEPs can be obtained relatively easily; they are noninvasive and require 30 to 40 minutes to perform. Short-latency SEPs reflect the subcortical and primary cortical aspects of the somatosensory pathways. They are not affected by the sedative and analgesic drugs used in intensive care 9-1~ and can be used in patients receiving muscle relaxants, in whom clinical assessment is impossible. Studies have been done on auditory and visual evoked potentials as predictors of outcome, but these indicators do not add further prognostic information; both may be unaffected by brain injury so that, although loss of either is associated with poor outcome, neither is reported to be as sensitive as SEPs. 1, 2, 7, 12 It has been proposed that evoked potentials be used as an adjunct in planning therapy13; Judson et al. 5 have incorporated SEPs into the management of head injury in adults. There are only a few studies of SEPs in children with brain injuries,2, 6, 7, 14 and no studies, of either pediatric or adult patients, provide details of confidence limits. The aims of this study were to evaluate SEPs prospectively as a predictor of neurologic outcome in a large number of children with acute, severe brain injury. Our hypothesis was that no child would have a good recovery if SEPs were absent bilaterally at any time. If this generalization was not true, we wished to determine whether exceptions could be defined--in particular, whether focal intracranial abnormalities could lead to loss of SEPs but the child could still make a good recovery. We also compared the predictive power of SEPs with clinical assessment and, in patients with traumatic brain injury, with computed tomographic scan. METHODS Patients. Children eligible for this study were those older than 1 month of age with severe brain injury who were admitted to the intensive care unit at the Hospital for Sick Children (HSC), Toronto, Canada, from July 1990 to February 1992, and at the Royal Children's Hospital (RCH), Melbourne, Australia, from February 1991 to February 1992. Severe brain injury was defined as (1) a Glasgow Coma Scale score <8, (2) the inability to localize painful stimuli if a full Glasgow Coma Scale was not performed (i.e., motor score <4), or (3) a history suggestive of severe brain injury in a patient who could not be examined neurologically. All patients had endotracheal tubes in place and were supported by mechanical ventilation. Intracranial
45
pressure was monitored in children with traumatic brain injury. Patients were excluded if they were known to have neurological deficits before the acute event, or if they subsequently died from nonneurologic causes, or if a diagnosis of brain death was made before SEPs were performed. We performed SEPs in 120 patients with acute traumatic and nontraumatic brain injury. Of these, 11 were excluded from further analysis either because they died from nonneurologic causes (n = 8) or had neurodevelopmental disabilities before the injury (n --- 3). The median age of the remaining 109 patients was 4.8 years (range, 2 months to 16.8 yrs): Infants younger then 1 month of age were not included because of the requirement of some history of normal neurologic development before injury (for our criteria), and we did not wish to include the range of possible perinatal injuries. The application of evoked potentials in both term and preterm infants warrants separate studies. Somatosensory evoked potentials can be reliably recorded in all normal infants if measurements used for older children are modified for those younger than 4 months 15, 16; if these modifications are not made, this is not the case. 17, 18 In our study there were only two children younger than 4 months of age (both were 2 months of age). The mechanisms of brain injury were trauma in 53 (48.6%) patients, hypoxic ischemic encephalopathy in 33 (30.3%), and other in 23 (21.1%). The latter group included meningitis (n = 6), hemorrhage (n = 3), tumor (n = 3), cardiopulmonary bypass (n = 3), hemolytic uremic syndrome (n = 2), insulin overdose (n = 1), hepatic encephalopathy (n = 1), and encephalopathy of unknown cause (n = 4). Children with acute brain dysfunction after cardiopulmonary bypass were classified as "other" because of the several possible mechanisms of injury. The Glasgow Coma Scale score or the motor score alone was recorded at the time of admission. If the patient was unable to be examined clinically because of the effects of muscle relaxants, the most recent assessment before admission was used (for example, by the referring hospital or an ambulance attendant). All patients had CT scans as part of their diagnostic evaluation. The first CT scan after admission was classified according to the Traumatic Coma Data Bank scale for patients with traumatic brain injury. 19 The outcome was assessed with the Glasgow Outcome Scale 2° at a minimum of 6 months after injury. This scale consists of five categories of patient outcomes: good (normal or mild disability), moderate (disabled but independent), severe (conscious but severely disabled and dependent), persistent vegetative state, and death. For analysis these groups were further simplified into favorable outcome (including good or moderate disability) and unfavorable outcome (including severe disability, vegetative state, or death). Outcome assessment was performed by pediatric neurologists with no knowledge of the results of the SEPs.
46
Beca etal.
The Journal of Pediatrics January 1995
Table I. Result and outcome of SEPs
and the worst SEPs recorded for each child. Confidence intervals were derived with the binomial distribution.23
Outcome
Initial SEPs Normal Unilateral delay Bilateral delay Unilateral absent Bilateral absent Worst SEPs Normal Unilateral delay Bilateral delay Unilateral absent Bilateral absent
Favorable* (No.)
Unfavorablet (No.)
31 4 5 9 1
3 0 5 10 41
26 5 4 11 4
2 0 4 8 45
*Normal or moderate disability. tSevere disability,vegetativestate, or death.
Somatosensory evoked potentials. The SEPs were recorded with a model 2000 measuring unit (Nicolet Instruments, Madison, Wis.) (HSC) or a Medilec Maestro measuring unit (Medilec, Old Woking, Surrey, England) (RCH). The right and left median nerves were stimulated at the wrist, and SEPs were recorded over the cervical spine (C7) and contralateral somatosensory cortex (C3' and Ca') with a frontal reference electrode (Fpz). If the wrist was inaccessible, stimulation was performed at the antecubital fossa. If the cervical spine was inaccessible, peripheral SEPs were recorded at the Erb point (over the brachial plexus). At least two averages of 256 artifact-free responses were recorded per arm. The bandpass was 30 to 3000 Hz (HSC) and 20 to 2000 Hz (RCH), the sweep interval was 50 ms, and the stimulation rate was 4.1 Hz. In infants younger than 4 months of age, 64 trials were averaged; the bandpass was 30 to 3000 Hz and 5 to 1500 Hz (HSC), or 20 to 2000 Hz and 3 to 1000 Hz (RCH), with a sweep of 200 ms, and a stimulation rate of 1.1 Hz. 19 The SEPs were classified into five categories: (1) normal, (2) unilateral increased latency, (3) bilateral increased latency, (4) unilateral absence, and (5) bilateral absence. Latency was considered prolonged if it was greater than 2.5 SD above normal for the child's age.iS, z1The first three categories were further subdivided into normal and low amplitude; category 4 was subdivided according to whether the unilateral response was normal or delayed. If there was uncertainty about classification, SEPs were recorded as the better alternative. When possible, SEPs were recorded within 48 hours of admission, and in all cases they were recorded within 4 days of the onset of coma. They were repeated within the first week in 67 of the 109 patients. The positive and negative predictive values, and the sensitivity and specificityz2 were calculated for the first SEPs
RESULTS For 80% of the patients, initial SEPs were performed within 48 hours, and for 91% within 72 hours; 54% of patients had SEPs repeated one to three times during the first week. The overall outcome was good in 34 (31.2%) of the 109 children, and there was moderate disability in 16 (14.7%), severe disability in 13 (11.9%), persistent vegetative state in 9 (8.3%), and death in 37 (33.9%), yielding an overall favorable outcome of 44.9%. Reduced SEP amplitude did not appear to be predictive of outcome, but numbers within these subgroups were too small for meaningful analysis. Favorable outcome was associated with normal (category 1) SEPs at initial testing in 31 of 34 children (91% positive predictive power) (Table I). If the worst SEP recording was considered, the positive predictive power increased to 93% (95% CI, 77% to 99%). Of 42 children with bilaterally absent SEPs (category 5) at initial SEP testing, 41 had an unfavorable outcome (98% positive predictive power). This included a single child in whom SEPs were bilaterally absent for anatomic reasons (subdural effusions). In 49 children, SEPs were absent in the worst recording. Of these, 45 had an unfavorable outcome. (92% positive predictive power; 95% CI, 80% to 98%); 4 of these 45 children had an evoked response (categories 1 to 4) present on the initial recording that became absent on later testing. The sensitivity, specificity, and predictive value of later versus initial SEPs were compared; minimal differences were found (Table II). An additional four patients had bilaterally absent SEPs (category 5) at either initial or after additional testing and a favorable outcome (normal or moderate disability) (Table I). Two of these four children had meningitis with subdural effusions; one had a favorable outcome and one had moderate disability at follow-up. Of the remaining two children, one had traumatic brain injury with hemorrhage in the tegmentum of the mesencephalon and the right cerebral peduncle and contusion to the posterior limb of the left internal capsule; she had a favorable outcome with moderate disability. The fourth patient also had traumatic brain injury and grade I SEPs on the first recording; the SEPs became absent after decompressive craniectomy for uncontrolled intracranial hypertension, and then returned by day 7; she also had moderate disability. Each of these four children had identifiable clinical features known to interfere technically with the recording of SEPs. If these four patients are excluded, the positive predictive value for absent SEPs is 100% (95% CI, 92% to 100%). Of the 45 children with bilaterally absent SEPs (category 5) on either initial or subsequent testing and poor outcome, 27% were alive at 6 months, but were either
The Journal o f Pediatrics Volume 126, Number 1
Beca et al.
47
Table II. Predictive value, sensitivity, and specificity for SEPs and motor score
Worst SEPs Absent (category 5)/unfavorable Normal (category 1)/favorable Initial SEPs Absent (category 5)/unfavorable Normal (category l)/favorable Motor score Flaccid/unfavorable
Positive predictive value
Negative predictive value
Sensitivity
Specificity
45/49 (92) 26/28 (93)
46/60 (77) 57/81 (70)
45/59 (76) 26/50 (52)
46/50 (92) 57/59 (97)
41/42 (98) 31/34 (91)
49/67 (73) 56/75 (75)
41/59 (69) 31/50 (62)
49/50 (98) 56/59 (95)
21/21 (100) 38/63 (60) 21/46 (46) Data are presentedas numberwith test result/numberwith outcome,Numbersin parenthesesindicatepercentages. severely disabled or in a persistent vegetative state. Of the 32 patients with abnormal but present SEPs (categories 2 to 4), 20 (63%) had favorable outcomes. All the patients with no response to pain had an unfavorable outcome (Table III). Of the 32 patients with flexor or extensor posturing responses, 14 (44%) had a favorable outcome, and 24 (77%) of 31 with withdrawal or a localizing response had a favorable outcome. The motor score could not be recorded in 25 (22.9%) of the patients, usually because muscle relaxants or large doses of sedative drugs were used, but some motor assessments were either incorrectly performed or not recorded by the admitting physicians. An absent motor response to painful stimulus had a positive predictive power of 100% (95% CI, 84% to 100%) for unfavorable outcome. However, sensitivity was 46% in those who could be clinically assessed and only 36% in all patients, compared with sensitivity of 69% for initial SEPs. No specific CT scan appearance in patients with traumatic brain injury was invariably associated with either a good or bad outcome (Table IV); 38 (72%) of the 53 patients had diffuse injury II or III (Traumatic Coma Data Bank scale), and in these groups CT scan had poor predictive power. Numbers in other categories were very small. DISCUSSION This study shows that bilaterally absent SEPs accurately predict unfavorable outcome in the setting of both traumatic and nontraumatic coma. In any test of neurologic function that attempts to predict neurologic outcome, the positive predictive power and the specificity are of much greater importance than the sensitivity. It would be highly undesirable for treatment to be withdrawn because a test incorrectly predicted a poor outcome, but less unacceptable if treatment were continued because a test incorrectly predicted a good outcome. Knowledge of the CIs is essential, to proper test use, but in published studies on SEPs these have not been provided. For a test to be clinically useful, it is necessary to know the outcome for a large number of patients with thepoor prognostic sign.24, 25The clinician needs
38/38 (100)
Table III. Motor score related to outcome: Number of children (percent) Outcome Favorable
Motor score Localizes Withdrawal Flexor Extensor Flaccid Inaccessible
Unfavorable
No.
%
No.
%
6 18 7 7 0 12
75 78 41 47 0 48
2 5 10 8 21 13
25 22 59 53 100 52
Table IV. Computed tomographic scan related to outcome of traumatic brain injury Outcome Favorable
Diffuse injury I (normal) II (minor abnormality) III (edema) IV (shift) Evacuated mass Nonevacuated mass
Unfavorable
No.
%
No.
%
1 9 12 1 5 1
50 60 52 25 83 33
I 6 11 3 1 2
50 40 48 75 17 67
to know what proportion of children with absent SEPs will have a poor outcome. In this study initial SEPs were only slightly different than additional SEPs performed later. Although many children had SEPs that showed some improvement or deterioration in latency and amplitude, only a few children had additional SEPs that crossed the present/absent boundaries. A larger series is needed to determine the significanceof the changes, such as from abnormal to normal latencies. Normal SEPs, regardless of when recorded, predicted good outcome in 91% to 93% of children; abnormal but present responses
48
Beca et al.
predicted good outcome in 63%. Absent SEPs predicted unfavorable outcome in 92% of children. Absent SEPs on later studies in patients with no identifiable anatomic cause for their absence always predicted poor outcome (95% CI, 92% to 100%). If the lower limit of the 95% CI is used, this study shows that up to 8% of children with absent SEPs could still have a favorable outcome. However, given the very high predictive power shown in many adult and a few pediatric studies, 13 the risk must be much lower than this. With Bayesian methods, the actual risk is likely to be 2% for this sample sizeY Huge numbers would be required to reduce the lower limit of the 95% CI to less than 1%, and ultimately the decision about how to use this information will depend on the individual doctor, hospital, and country. Although we have shown that absent SEPs have high sensitivity and specificity, in certain instances they cannot be relied on to predict outcome accurately; these include extraaxial collections, brain stem hemorrhage, and after cranial decompression. Extraaxial collections, such as subdural effusions in meningitis, increase the physical distance between the brain and recording electrodes, which can be sufficient to prevent the recording of responses. However, because CT scanning is part of the management of such conditions, the presence of these collections will be known at the time. Likewise, CT scanning will reveal brain stem hemorrhage, which can interfere with conduction in the subcortical somatosensory pathways. Deep cerebral hemorrhages in traumatic brain injury, especially if in the brain stem, are in themselves associated with poor outcome. Midbrain hemorrhage in adults has been shown to lead to loss of SEPs and has been uniformly associated with poor outcome.26, 27 However, because the child with brain stem hemorrhage in this study survived with moderate disability, we suggest that absent SEPs should not be used to predict outcome in cases with brain stem hemorrhage. The final exception, of cranial decompression for intracranial hypertension, has been previously reported in children with Reye syndrome.2 Taylor and Farrell7 also identified a child with brain injury as a result of a lightning strike who had transient loss of SEPs and good outcome. Both these clinical situations are encountered so rarely that it will be difficult to get further information, but in the former, SEPs recorded before decompression accurately predicted the outcome. If these conditions are considered, we have shown that absent SEPs are relatively specific in predicting poor outcome. It is essential that technically satisfactory SEPs be obtained, and recordings in which the components are obscured by artifact should not be used clinically. We have found that electrical interference is rarely a problem, even in an intensive care setting. The most common source of artifact is from muscle activity in nonparalyzed patients. Clear cervical or brachial plexus potentials, demonstrating
The Journal of Pediatrics January 1995
the integrity of the peripheral nervous system, must be obtained before absent cortical potentials are regarded as significant. Absence of motor response to painful stimuli was as specific as bilaterally absent SEPs for predicting unfavorable outcome but considerably less sensitive. In evaluating clinical assessment in this study, motor response was used for two reasons. First, it is likely that most, if not all, the predictive power of the Glasgow Coma Scale is derived from the motor score. 28 Moreover, patients with nontraumatic coma commonly had only the motor score rather than the fuli Glasgow Coma Scale score recorded. Muscle relaxants and sedative and analgesic agents are frequently used in neurointensive care; thus it is often difficult to obtain additional clinical assessments of patients with acute brain injury, but SEPs can be used in these patients. No CT scan appe~trance had absolute specificity, although assessment was limited by the small number of patients in the several subgroups. The distinction between moderate and severe disability at clinical follow-up is difficult in children, and becomes more difficult with younger age; this distinction is critical because we used it to separate favorable from unfavorable outcome in this study. However, all our patients with severe disabilities had profound developmental delay (functioning at less than half the appropriate level for age) and severe motor disabilities (preventing independent mobility). We suggest that SEPs be performed in all children with acute, severe brain injury, because they are the best predictor of outcome in this group. The SEPs should be performed within 24 to 48 hours of admission and be repeated within 24 to 48 hours to increase positive predictive power. If a CT scan excludes th6 presence of extraaxial fluid collections or brain stem hemorrhage, the presence of normal SEPs suggests a 93% chance of a favorable outcome, whereas the absence of SEPs predicts a 92% chance of an unfavorable outcome. In patients who had no physical barrier impeding the recording of the cortical response, absent SEPs predicted a 100% chance of unfavorable outcome. If SEPs are absent and clinical assessment is compatible with an unfavorable outcome, we believe that consideration should be given to withdrawal of treatment. We thank L. J. MacMillan, N. K. Keenan, and B. G. Carter for their excellent technical support. REFERENCES
1. De Meirleir DJ, Taylor MJ. Evoked potentials in comatose children: auditory brainstem responses. Pediatr Neurol 1986; 2:31-4. 2. De Meirleir LJ, Taylor MJ. Prognostic utility of SEPs in comatose children. Pediatr Neurol 1987;3:78-82. 3. Greenberg RP, Newlon PG, Becker DP. The somatosensory
The Journal o f Pediatrics Volume 126, Number 1
4. 5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
Beca et al.
evoked potential in patients with severe head injury: outcome prediction and monitoring of brain function. Ann N Y Acad Sci 1982;388:683-8. Hume AL, Cant BR. Central somatosensory conduction after head injury. Ann Neurol 1981;10:411-9. Judson JA, Cant BR, Shaw NA. Early prediction of outcome from cerebral trauma by somatosensory evoked potentials. Crit Care Med 1990;18:363-8. Lutschg J, Pfenninger J, Ludin HP, Vassella F. Brain-stem auditory evoked potentials and early somatosensory evoked potentials in neurointensively treated comatose children. Am J Dis Child 1983;137:421-6. Taylor M J, Farrell EJ. Comparison of the prognostic utility of VEPs and SEPs in comatose children. Pediatr Neurol 1989; 5:145-50. Taylor M J, Murphy W J, Whyte HE. Prognostic reliability of somatosensory and visual evoked potentials of asphyxiated term infants. Dev Med Child Neurol 1992;34:507-15. Ganes T, Lundar T. The effect of thiopentone on somatosensory evoked responses and EEGs in comatose patients. J Neurol Neurosurg Psychiatry 1983;46:509-14. Newlon PG, Greenberg RP, Hyatt MS, Enas GG, Becket DP. The dynamics of neuronal dysfunction and recovery following severe head injury assessed with serial multimodality evoked potentials. J Neurosurg 1982;57:168-77. Sutton LN, Frewen T, Marsh R, Jaggi J, Bruce DA. The effects of deep barbiturate coma on multimodality evoked potentials. J Neurosurg 1982;57:178-85. Cant BR, Hume AL, Judson JA, Shaw NA. The assessment of severe head injury by short-latency somatosensory and brain-stem auditory evoked potentials. Electroencephalogr Clin Neurophysiol 1986;65:188-95. Goodwin SR, Friedman WA, Bellefteur M. Is it time to use evoked potentials to predict outcome in comatose children and adults? Crit Care Med 1991;19:518-24. Frank LM, Furgiuele TL, Etheridge JE Jr. Prediction of chronic vegetative state in children using evoked potentials. Neurology 1985;35:931-4.
49
15. George SR, Taylor MJ. Somatosensory evoked potentials in neonates and infants: developmental and normative data. Electroencephalogr Clin Neurophysiol 1991 ;80:94-102. 16. Bongers-Schokking C J, Colon E J, Hoogland RA, Van den Brande JL, De Groot CJ. The somatosensory evoked potentials of normal infants: influence of filter bandpass, arousal state and number of stimuli. Brain Dev 1989; 11:33-9. 17. Willis J, Seales D, Frazier B. Short latency somatosensory evoked potentials in infants. Electroencephalogr Clin Neurophysiol 1984;59:366-73. 18. Laureau E, Majnemer A, Rosenblatt B, Riley P. A longitudinal study of short latency somatosensory evoked responses in healthy newborns and infants. Electroencephalogr Clin Neurophysiol 1988;71:100-8. 19. Marshall LF, Marshall SB, Klauber MR, et al. A new classification of head injury based on computerised tomography. J Neurosurg 1991;75:S14-20. 20. Jennett B, Bond M. Assessment of outcome after severe brain damage. Lancet 1975;1:480-4. 21. Taylor M J, Fagan ER. SEPs to median nerve stimulation: normative data for paediatrics. Electroencephalogr Clin Neurophysiol 1988;71:323-30. 22. Altman DG. Practical statistics for medical research. 1st ed. London: Chapman and Hall, 1990:409-16. 23. Lentner C. Geigy scientific tables, basel: Ciba Geigy, 1982:89102. 24. Bates D. Defining prognosis in medical coma. J Neurol Neurosurg Psychiatry 1991;54:569-71. 25. Shewmon DA, De Giorgio CM. Early prognosis in anoxic coma: reliability and rationale. Neurol Clin 1989;7:823-43. 26. Rosenblum WI, Greenberg RP, Seelig JM, Becker DP. Midbrain lesions: frequent and significant prognostic feature in closed heart injurt. Neurosurgery 1981;9:613-20. 27. Marshall LF, Gautille T, Klauber MR, et al. The outcome of severe closed head injury. J Neurosurg 1991;75:$28-36. 28. Jagger J, Jane JA, Rimel R. The Glasgow Coma Scale: To sum or not to sum? Lancet 1983;2:97.
AVAILABILITY OF JOURNAL BACK ISSUES As a service to our subscribers, copies of b a c k issues of THE JOURNAL OF PEDIATRICS for the preceding 5 years are m a i n t a i n e d and are available for purchase from the publisher, M o s b y - Y e a r Book, Inc., at a cost of $10.00 per issue. T h e following quantity discounts are available: 25% off on quantities of 12 to 23, and one third off on quantities of 24 or more. Please write to M o s b y Year Book, Inc., Subscription Services, 11830 Westline Industrial Drive, St. Louis M O 631463318, or call (800)453-4351 or (314)453-4351 for information on availability of particular issues. If back issues are unavailable from the publisher, photocopies of complete issues are available from U M I , 300 N. Z e e b Rd., A n n Arbor, M I 48106, (313)761-4700.