- Email: [email protected]
Factors associated with adverse outcomes in children with diabetic ketoacidosis-related cerebral edema
Factors associated with adverse outcomes in children with diabetic ketoacidosis-related cerebral edema James P. Marcin, MD, MPH, Nicole Glaser, MD, Peter Barnett, MB, BS, Ian McCaslin, MD, David Nelson, MD, Jennifer Trainor, MD, Jeffrey Louie, MD, Francine Kaufman, MD, Kimberly Quayle, MD, Mark Roback, MD, Richard Malley, MD, and Nathan Kuppermann, MD, MPH, for The Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pediatrics Objective: To investigate the relation between outcomes of children with diabetic ketoacidosis (DKA)-related cerebral edema and baseline clinical features and therapeutic interventions for treatment of cerebral edema. Study design: All children ≤18 years old with DKA and cerebral edema (n = 61) were retrospectively identified from 10 pediatric centers between 1982 and 1997. Demographic, biochemical, and therapeutic data were collected. Ordinal logistic regression analysis was used to identify factors associated with the clinical outcome (death or persistent vegetative state; mild to moderate neurological disability; or normal) after adjusting for known risk factors for the development of cerebral edema as well as the degree of neurologic depression at the time of diagnosis of cerebral edema. Results: Seventeen (28%) children died or survived in a vegetative state; 8 (13%) survived with mild to moderate neurologic disabilities; and 36 (59%) survived without sequelae. Factors associated with poor outcomes included greater neurologic depression at the time of diagnosis of cerebral edema, a high initial serum urea nitrogen concentration, and intubation with hyperventilation to a PCO2 <22 mm Hg. Conclusions: After adjusting for potential confounding variables and the degree of neurologic compromise at the initiation of therapy, intubation with hyperventilation is associated with adverse outcomes of DKA-related cerebral edema. Greater neurologic depression at the time of diagnosis of cerebral edema and a higher initial serum urea nitrogen concentration are also associated with poor outcome. (J Pediatr 2002;141:793-97) From the Department of Pediatrics and the Division of Emergency Medicine, Department of Internal Medicine, University of California, Davis, School of Medicine, Davis; the Division of Emergency Medicine, Children’s Hospital and Health Center, San Diego; the Department of Pediatrics, Children’s Hospital of Los Angeles, University of Southern California School of Medicine, Los Angeles, California; the Division of Emergency Medicine, Royal Children’s Hospital, Melbourne, Australia; the Department of Pediatrics, Brown University School of Medicine, Providence, Rhode Island; the Division of Emergency Medicine, Children’s Memorial Hospital, Chicago, Illinois; the Division of Emergency Medicine, Children’s Hospital of Philadelphia, Pennsylvania; the Department of Pediatrics, Washington University School of Medicine, St Louis, Missouri; the Department of Pediatrics, University of Colorado Health Sciences Center, Denver, Colorado; and the Department of Pediatrics, Harvard Medical School, Boston, Massachusetts.
Submitted for publication Dec 28, 2001; revisions received Mar 27, 2002, and June 12, 2002; accepted July 31, 2002. Reprint requests: James P. Marcin, MD, MPH, Section of Pediatric Critical Care, University of California, Davis, Children’s Hospital, 2516 Stockton Blvd, Sacramento, CA 95817. Copyright © 2002, Mosby, Inc. All rights reserved. 0022-3476/2002/$35.00 + 0 9/21/128888 doi:10.1067/mpd.2002.128888
Diabetic ketoacidosis (DKA) in pediatric patients is a relatively common indication for admission to the pediatric intensive care unit (PICU). In a cohort of patients admitted to 32 PICUs across the United States, DKA is noted as the primary acute diagnosis necessitating admission in 2.4% of ad-
See editorial, p 754. missions.1 In addition to the need for intensive physician and nursing care, frequent laboratory analyses, and the risk of serious biochemical abnormalities, pediatric patients with DKA may require admission to the PICU because of concerns regarding the risk of cerebral edema. This complication occurs in ~1% of children with DKA and has a reported mortality rate from 21% to 90%.2-7 BUN DKA GCS PICU PVS
Blood urea nitrogen Diabetic ketoacidosis Glasgow Coma Score Pediatric intensive care unit Persistent vegetative state
There are many published treatment recommendations for children with DKA directed at minimizing the risk of cerebral edema.3,5,7-10 These recommendations are generally based on retrospective reviews of associated clinical and laboratory factors. There has been little research, however, to determine the optimal therapy for DKA-related cerebral edema once it is suspected or diagnosed. In this study, we investigated factors associated with outcome in a nonconcurrent cohort of children with 793
MARCIN ET AL
THE JOURNAL OF PEDIATRICS DECEMBER 2002
Table I. Neurologic symptom score
Score
Clinical descriptions
1 2 3 4 5
Irritable; disoriented; confused; or GCS = 13–15 Lethargic; somnolent; or GCS = 11–12 Stuporous; purposeful response to pain; or GCS = 8–10 Abnormal or absent purposeful response to pain; or GCS = 6-7 Focal neurologic finding; fixed and dilated pupil(s); respiratory arrest; or GCS = 3-5
DKA-related cerebral edema. The factors analyzed included biochemical abnormalities at presentation and during therapy for DKA, neurologic symptoms and signs at the time of diagnosis of cerebral edema, and therapies administered to treat cerebral edema.
METHODS Patients in this study were identified in a multicenter collaborative research study conducted through the Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pediatrics. Methods regarding patient identification and data collection have previously been published.7 Patients 18 years of age or younger with DKA and cerebral edema examined at 10 pediatric centers between 1982 and 1997 were identified. Cerebral edema was defined as an alteration in mental status (obtundation or disorientation) and either radiographically or pathologically confirmed cerebral edema or clinical improvement after a specific treatment for cerebral edema. The primary clinical outcome for this study was categorized according to survival and discharge neurologic status as follows: death or persistent vegetative state (PVS)11,12; survival with mild to moderate neurologic disability; and survival with normal neurologic outcome. Outcomes were determined by review of the medical records by two physicians who were blinded to all potential independent variables, including hospital identification and therapies for treatment. Of the 794
61 patients, there was disagreement only in two outcome assignments (κ = 0.86), which were resolved by consensus among three of the investigators. Two groups of independent variables were included in the multivariate analyses. The first group consisted of variables that were previously identified as independent risk factors for the development of cerebral edema,7 some of which reflect the severity of illness: the initial partial pressure of arterial carbon dioxide, the initial blood urea nitrogen (BUN) concentration, treatment with bicarbonate, and changes in measured (uncorrected) serum sodium concentration during therapy for DKA. In addition, to adjust for the severity of neurologic depression at the time of diagnosis of cerebral edema, a 5-category neurologic symptom score was developed (Table I). Because many patients did not have a Glasgow Coma Score (GCS) recorded at the time of neurologic deterioration, the neurologic symptom score was designed as an approximation of the GCS. This scoring system was defined a priori by the research team before review of patient data. Each of the patients was assigned a neurologic symptom score by consensus among three of the investigators, based on the descriptions in the medical record of neurologic abnormalities and/or the GCS at the time of diagnosis of cerebral edema. The authors were blinded to the outcome when scoring patients. The second group of independent variables considered in the analyses consisted of therapeutic interventions directed at treating cerebral edema.
These included hyperosmolar therapies, glucocorticoid therapy, diuretic therapies, bicarbonate, intubation, and intubation with associated hyperventilation. Glucocorticoid therapies were standardized to dexamethasone, based on their glucocorticoid potency. The degree of hyperventilation was determined by the average arterial PCO2 within the first 3 hours after intubation. Patients were considered to have received a particular therapeutic intervention if the therapy was initiated or administered within 1 hour of diagnosis of cerebral edema. Continuous and ordinal independent variables were first analyzed with logit plots to assess their bivariate association with the outcomes. Transformations were made when appropriate. The logit analysis of hyperventilation and arterial PCO2 versus the log odds of death or a discharge diagnosis of PVS suggested a dichotomous relation best delineated at a PCO2 of 22 mm Hg. We therefore defined hyperventilation as intubation with a mean PCO2 during the first 3 hours after intubation <22 mm Hg. To evaluate the association of clinical outcome with baseline disease severity indicators and therapeutic interventions, we conducted an ordinal logistic regression analysis for the 3-category neurologic outcome variable. Validation was performed by means of bootstrap analysis with 1000 iterations with selection of variables at a P value of .05. Variables were considered significantly associated with the outcome if selected in >50% of the bootstrap ordinal logistic regression iterations.13 All of the statistical computations were 2-tailed and were performed with Stata statistical software (Version 7.0, Stata, College Station, Tex).
RESULTS Of 61 patients identified as having cerebral edema, 13 (21%) died and 4 (7%) survived in a PVS; 36 (59%) patients survived with no persistent neu-
MARCIN ET AL
THE JOURNAL OF PEDIATRICS
VOLUME 141, NUMBER 6 rologic deficits, and 8 (13%) children survived with mild or moderate persistent neurologic disabilities. The age distribution of the children diagnosed with cerebral edema is shown in Fig 1. Three patients were diagnosed with cerebral edema at the time of initial presentation to the emergency department. In the remainder, cerebral edema was diagnosed between 1 and 25 hours after the initiation of therapy for DKA (Fig 2). There was no correlation of clinical outcome with the time of diagnosis of cerebral edema during therapy (Kruskal-Wallis, P = .12). The distribution of the neurologic symptom score and clinical outcome is shown in Fig 3. Of note, all children who either died or survived in a PVS had a neurologic symptom score of 4 or 5 at the time of diagnosis of cerebral edema. Therapies initiated within 1 hour of the diagnosis of cerebral edema are shown in Table II. Mannitol was the most commonly used therapy (n = 41, 67%). Approximately half of the patients were intubated within 1 hour of diagnosis of cerebral edema (n = 33, 54%), but only 17 (28%) had associated hyperventilation (PCO2 <22 mm Hg). Other therapies (glucocorticoids, bicarbonate, and furosemide) were used infrequently and therefore were not included in the multivariate analysis. The results of the ordinal logistic regression are shown in Table III. Variables significantly associated with poor outcomes include a higher neurologic symptom score at the time of diagnosis of cerebral edema, a higher initial BUN concentration, and intubation with associated hyperventilation. In addition, there was a trend toward an association between poor clinical outcome and a lesser rise in measured serum sodium concentration during therapy for DKA. No other variables were significantly associated with clinical outcome. In a subanalysis, we included intubation alone (with or without associated hyperventilation) in place of intubation with hyperventilation in the regression equation. In this subanalysis, intubation alone was
Table II. Therapies initiated within 1 hour of diagnosis of cerebral edema
Outcome
Therapy
Total n = 61
Normal n = 36
Mild-tomoderate disability n=8
Hyperosmolar therapies Mannitol 3% Saline Glucocorticoid therapy Furosemide Bicarbonate Intubation Intubation with hyperventilation
41 (67) 1 (2) 2 (3) 8 (13) 2 (3) 33 (54) 17 (28)
23 (64 1 (3) 1 (3) 1 (3) 1 (3) 13 (36) 4 (11)
5 (63)) 0 (0) 0 (0) 3 (38) 0 (0) 6 (75) 2 (25)
Death or PVS n = 17 13 (76) 0 (0) 1 (6) 4 (24) 1 (6) 14 (82) 11 (65)
Data are given as n (%).
not associated with worse outcomes after adjusting for the other variables (data not shown). The bootstrap analysis was consistent with the ordinal logistic regression analysis. Of the 1000 bootstrap iterations, the same 3 variables (neurologic symptom score, BUN, and intubation with hyperventilation) were selected as significantly associated with poorer clinical outcomes in >50% of the iterations of the logistic regression (Table III).
DISCUSSION In the current study, 3 variables were found to be associated with a poor outcome of children with DKA-related cerebral edema. These variables included an elevated initial BUN concentration, more profound neurologic depression at the time of diagnosis of cerebral edema, and intubation with associated hyperventilation to a PCO2 level <22 mm Hg. Because the neurologic symptom score is a marker for the severity of neurologic abnormalities at the time of diagnosis of cerebral edema, its association with poor outcome is not surprising. The degree of neurologic impairment at the time of diagnosis of cerebral edema has been noted in a pre-
vious study to be correlated with outcome.5 In that study, the author found that children who had had a respiratory arrest at the time of diagnosis of cerebral edema almost uniformly died or survived with severe neurologic deficits.5 The association of elevated initial BUN concentration with poor outcome suggests that patients with more severe dehydration at initial examination are more likely to have adverse outcomes. Previous studies have demonstrated an association between the risk of DKA-related cerebral edema and the degree of initial dehydration.7,14 The findings of the current study therefore strengthen the evidence for a possible pathophysiologic association between dehydration (and possible cerebral ischemia) and cerebral edema in the setting of DKA. Intubation with associated hyperventilation was also associated with a poorer neurologic outcome. This result is consistent with research that has suggested that therapeutic hyperventilation can decrease cerebral blood flow and contribute to cerebral ischemia, potentially exacerbating evolving brain injury. In the setting of traumatic brain injury, the evidence linking hyperventilation and worse neurologic outcomes is growing.15-17 The largest published, 795
MARCIN ET AL
THE JOURNAL OF PEDIATRICS DECEMBER 2002
Fig 1. Age distribution of children diagnosed with DKA-related cerebral edema.
Fig 2. Distribution of clinical outcome and time between initiation of therapy for DKA and diagnosis of cerebral edema.
Table III. Associations of variables with adverse clinical outcome of DKA-related cerebral edema
Coefficient Age (y) –0.084 –0.0091 Initial arterial PCO2 –0.10 Change in serum sodium* Initial BUN concentration 0.086 0.70 Bicarbonate therapy† Neurologic symptom score 2.2 Mannitol therapy –0.053 Intubation and hyperventilation 2.1
95% CI –0.27-0.11 –0.13-0.11 –0.23-0.02 0.01-0.16 –1.13-2.53 1.06-3.37 –1.77-1.67 0.29-3.84
Bootstrap P value analysis (%)‡ .39 .88 .09 .02 .45 .001 .95 .02
7.0 5.1 28.7 59.2 9.6 95.2 9.3 55.3
*Represents change in measured serum sodium concentration during treatment of DKA before diagnosis of cerebral edema. †Administration of bicarbonate before diagnosis of cerebral edema. ‡Percentage of 1000 bootstrap iterations in which variable was selected as significantly associated with outcome.
prospective, randomized trial comparing hyperventilation (PCO2 25 ± 2 mm Hg) versus normoventilation (PCO2 35 ± 2 mm Hg) for acute traumatic brain injury demonstrated that hyperventilation resulted in a significant reduction in the number of patients with a favorable outcome (good or moderately disabled) at 3 and 6 months after injury.18 On the basis of these and other retrospective and animal data, the American Association of Neurological Surgeons has recommended that prophylactic hyperventilation be avoided in the setting of traumatic brain injury.19 Current reviews and texts suggest that in children with DKA-related cerebral edema, prompt control of the airway and ventilation as well as administration of mannitol are the primary therapies,20-22 and glucocorticoid med796
ications are thought not to be effective. These recommendations, however, are based mainly on case reports, because no previous study has investigated the effects of specific therapeutic interventions for this condition.22-24 In the current study, the use of mannitol was not associated with improved outcomes. These findings are similar to those of other studies investigating the use of mannitol in patients with cerebral edema of causes other than DKA, including traumatic brain injury,17,25 stroke,26,27 and hypoxic-ischemic injuries.21,28 In the current study, however, the dosages of mannitol differed among patients (dosages ranged from 0.2 g/kg body wt to 2.0 g/kg). In addition, although all patients categorized as having received mannitol were treated within 1 hour of the diagnosis of cere-
Fig 3. Distribution of clinical outcome and neurologic symptom score. bral edema, the timing of administration in relation to the earliest symptoms of cerebral edema was not known and probably differed substantially among patients. Furthermore, because of the small sample size, our ability to detect some clinically important associations between predictive variables and outcome was potentially limited. It is therefore difficult to draw any firm conclusions regarding the therapeutic efficacy of mannitol from this study. There are several other potential limitations to this study. First, because these analyses were performed on a retrospectively collected data set, the timing of therapeutic interventions and dosages of medications differed among patients. In addition, the administration of therapies was not randomized, so it is likely that therapeutic decisions were biased by disease severity, the time period during which the patient presented, and by physician and institutional preferences. Although we attempted to control for these potential biases by including variables related to disease severity in our analysis and by conducting a multivariate analysis, this bias nonetheless may not have been completely eliminated. In addition, some of the children who met the criteria for cerebral edema did not have the diagnosis confirmed radiographically or by postmortem examination. It is therefore possible that the neurologic abnormalities observed in some of the patients were due to other causes. Furthermore, the severity of neurologic depression at the time of diagnosis, as reflected in the
MARCIN ET AL
THE JOURNAL OF PEDIATRICS
VOLUME 141, NUMBER 6 neurologic symptom score, was determined through review of physician and nursing medical records. These data were recorded in varying degrees of detail and may not have accurately reflected the neurologic state of the patient at the time of diagnosis of cerebral edema. Finally, with regard to the association of intubation with associated hyperventilation and adverse outcome, it is unclear whether this association represents an adverse effect of treatment or possibly reflects the severity of the underlying illness. Although we attempted to control for illness severity and the degree of endogenous hyperventilation at presentation (ie, PCO2), it is nonetheless possible that the observed hyperventilation may reflect either a response to persistent acidosis or neurogenic hyperventilation in response to brain injury. We conclude that children with DKA-related cerebral edema who are most likely to have adverse neurologic outcomes are those who have greater degrees of dehydration as reflected by a higher BUN at the initiation of therapy for DKA and those who have more profound neurologic depression at the time of diagnosis of cerebral edema. Intubation with associated hyperventilation to a PCO2 <22 mm Hg is also correlated with a poorer neurologic outcome, although the current data do not allow us to definitively discern whether this association is an effect of therapy or possibly a reflection of the underlying disease state. Other therapies, including mannitol and intubation in general, were not significantly associated with worse or improved outcomes, although we were limited in our ability to study these associations. The effects of these therapies, in addition to the use of glucocorticoid medications, bicarbonate, and diuretics other than mannitol, warrant further investigation.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
REFERENCES 1. Marcin JP, Slonim AD, Pollack MM, Ruttimann UE. Long-stay patients in
the pediatric intensive care unit. Crit Care Med 2001;29:652-7. Bello F, Sotos J. Cerebral oedema in diabetic ketoacidosis in children. Lancet 1990;336:64. Duck S, Wyatt D. Factors associated with brain herniation in the treatment of diabetic ketoacidosis. J Pediatr 1988;113:10-4. Krane E, Rockoff M, Wallman J, Wolfsdorf J. Subclinical brain swelling in children during treatment of diabetic ketoacidosis. N Engl J Med 1985; 312:1147-51. Rosenbloom A. Intracerebral crises during treatment of diabetic ketoacidosis. Diabetes Care 1990;13: 22-33. Edge JA, Hawkins MM, Winter DL, Dunger DB. The risk and outcome of cerebral oedema developing during diabetic ketoacidosis. Arch Dis Child 2001;85:16-22. Glaser N, Barnett P, McCaslin I, Nelson D, Trainor J, Louie J, et al. Risk factors for cerebral edema in children with diabetic ketoacidosis. N Engl J Med 2001;344:264-9. Harris G, Fiordalisi I, Harris W, Mosovich L, Finberg L. Minimizing the risk of brain herniation during treatment of diabetic ketoacidemia: a retrospective and prospective study. J Pediatr 1990;117:22-31. Mahoney C, Vlcek B, Del Aguila M. Risk factors for developing brain herniation during diabetic ketoacidosis. Pediatr Neurol 1999;21:721-7. Edge JA. Management of diabetic ketoacidosis in childhood. Br J Hosp Med 1996;55:508-12. Multi-Society Task Force on PVS. Medical aspects of the persistent vegetative state (part 1). N Engl J Med 1994;330:1499-508. Multi- Society Task Force on PVS. Medical aspects of the persistent vegetative state (part 2). N Engl J Med 1994;330:1572-9. Chen C, George S. The bootstrap and identification of prognostic factors via Cox’s proportional hazards regression model. Stat Med 1985;4:39-46. Hammond P, Wallis S. Cerebral oedema in diabetic ketoacidosis: still puzzling—and often fatal. BMJ 1992; 305:203-4. Raichle ME, Posner JB, Plum F. Cerebral blood flow during and after hyperventilation. Arch Neurol 1970;23: 394-403.16. C a r m o n a Suazo JA, Maas AI, van den Brink WA, van Santbrink H, Steyerberg
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
EW, Avezaat CJ. CO2 reactivity and brain oxygen pressure monitoring in severe head injury. Crit Care Med 2000;28:3268-74. Schierhout G, Roberts I. Hyperventilation therapy for acute traumatic brain injury. Cochrane Database Syst Rev 2000;2. Muizelaar JP, Marmarou A, Ward JD, Kontos HA, Choi SC, Becker DP, et al. Adverse effects of prolonged hyperventilation in patients with severe head injury: a randomized clinical trial. J Neurosurg 1991;75: 731-9. Brain Trauma Foundation. The American Association of Neurological Surgeons. The Joint Section on Neurotrauma and Critical Care. Hyperventilation. J Neurotrauma 2000;17: 513-20. Rogers MC, Nichols DG, Ackerman AD. Textbook of pediatric intensive care. 3rd ed. Baltimore: Williams & Wilkins; 1996. Mujsce DJ, Towfighi J, Stern D, Vannucci RC. Mannitol therapy in perinatal hypoxic-ischemic brain damage in rats. Stroke 1990;21:1210-4. Roberts M, Slover R, Chase H. Diabetic ketoacidosis with intracerebral complications. Pediatr Diabetes 2001;2: 103-14. Franklin B, Liu J, Ginsberg-Fellner F. Cerebral edema and ophthalmoplegia reversed by mannitol in a new case of insulin-dependent diabetes mellitus. Pediatrics 1982;69:87-90. Shabbir N, Oberfield SE, Corrales R, Kairam R, Levine LS. Recovery from symptomatic brain swelling in diabetic ketoacidosis. Clin Pediatr 1992;31: 570-3. Roberts I, Schierhout G, Alderson P. Absence of evidence for the effectiveness of five interventions routinely used in the intensive care management of severe head injury: a systematic review. J Neurol Neurosurg Psychiatry 1998;65: 729-33. Bereczki D, Liu M, Prado GF, Fekete I. Cochrane report: a systematic review of mannitol therapy for acute ischemic stroke and cerebral parenchymal hemorrhage. Stroke. 2000;31: 2719-22. Bereczki D, Liu M, do Prado GF, Fekete I. Mannitol for acute stroke (Cochrane Review). Cochrane Database Syst Rev 2001;1. Modell JH. Drowning. N Engl J Med 1993;328:253-6. 797
Submitted for publication Dec 28, 2001; revisions received Mar 27, 2002, and June 12, 2002; accepted July 31, 2002. Reprint requests: James P. Marcin, MD, MPH, Section of Pediatric Critical Care, University of California, Davis, Children’s Hospital, 2516 Stockton Blvd, Sacramento, CA 95817. Copyright © 2002, Mosby, Inc. All rights reserved. 0022-3476/2002/$35.00 + 0 9/21/128888 doi:10.1067/mpd.2002.128888
Diabetic ketoacidosis (DKA) in pediatric patients is a relatively common indication for admission to the pediatric intensive care unit (PICU). In a cohort of patients admitted to 32 PICUs across the United States, DKA is noted as the primary acute diagnosis necessitating admission in 2.4% of ad-
See editorial, p 754. missions.1 In addition to the need for intensive physician and nursing care, frequent laboratory analyses, and the risk of serious biochemical abnormalities, pediatric patients with DKA may require admission to the PICU because of concerns regarding the risk of cerebral edema. This complication occurs in ~1% of children with DKA and has a reported mortality rate from 21% to 90%.2-7 BUN DKA GCS PICU PVS
Blood urea nitrogen Diabetic ketoacidosis Glasgow Coma Score Pediatric intensive care unit Persistent vegetative state
There are many published treatment recommendations for children with DKA directed at minimizing the risk of cerebral edema.3,5,7-10 These recommendations are generally based on retrospective reviews of associated clinical and laboratory factors. There has been little research, however, to determine the optimal therapy for DKA-related cerebral edema once it is suspected or diagnosed. In this study, we investigated factors associated with outcome in a nonconcurrent cohort of children with 793
MARCIN ET AL
THE JOURNAL OF PEDIATRICS DECEMBER 2002
Table I. Neurologic symptom score
Score
Clinical descriptions
1 2 3 4 5
Irritable; disoriented; confused; or GCS = 13–15 Lethargic; somnolent; or GCS = 11–12 Stuporous; purposeful response to pain; or GCS = 8–10 Abnormal or absent purposeful response to pain; or GCS = 6-7 Focal neurologic finding; fixed and dilated pupil(s); respiratory arrest; or GCS = 3-5
DKA-related cerebral edema. The factors analyzed included biochemical abnormalities at presentation and during therapy for DKA, neurologic symptoms and signs at the time of diagnosis of cerebral edema, and therapies administered to treat cerebral edema.
METHODS Patients in this study were identified in a multicenter collaborative research study conducted through the Pediatric Emergency Medicine Collaborative Research Committee of the American Academy of Pediatrics. Methods regarding patient identification and data collection have previously been published.7 Patients 18 years of age or younger with DKA and cerebral edema examined at 10 pediatric centers between 1982 and 1997 were identified. Cerebral edema was defined as an alteration in mental status (obtundation or disorientation) and either radiographically or pathologically confirmed cerebral edema or clinical improvement after a specific treatment for cerebral edema. The primary clinical outcome for this study was categorized according to survival and discharge neurologic status as follows: death or persistent vegetative state (PVS)11,12; survival with mild to moderate neurologic disability; and survival with normal neurologic outcome. Outcomes were determined by review of the medical records by two physicians who were blinded to all potential independent variables, including hospital identification and therapies for treatment. Of the 794
61 patients, there was disagreement only in two outcome assignments (κ = 0.86), which were resolved by consensus among three of the investigators. Two groups of independent variables were included in the multivariate analyses. The first group consisted of variables that were previously identified as independent risk factors for the development of cerebral edema,7 some of which reflect the severity of illness: the initial partial pressure of arterial carbon dioxide, the initial blood urea nitrogen (BUN) concentration, treatment with bicarbonate, and changes in measured (uncorrected) serum sodium concentration during therapy for DKA. In addition, to adjust for the severity of neurologic depression at the time of diagnosis of cerebral edema, a 5-category neurologic symptom score was developed (Table I). Because many patients did not have a Glasgow Coma Score (GCS) recorded at the time of neurologic deterioration, the neurologic symptom score was designed as an approximation of the GCS. This scoring system was defined a priori by the research team before review of patient data. Each of the patients was assigned a neurologic symptom score by consensus among three of the investigators, based on the descriptions in the medical record of neurologic abnormalities and/or the GCS at the time of diagnosis of cerebral edema. The authors were blinded to the outcome when scoring patients. The second group of independent variables considered in the analyses consisted of therapeutic interventions directed at treating cerebral edema.
These included hyperosmolar therapies, glucocorticoid therapy, diuretic therapies, bicarbonate, intubation, and intubation with associated hyperventilation. Glucocorticoid therapies were standardized to dexamethasone, based on their glucocorticoid potency. The degree of hyperventilation was determined by the average arterial PCO2 within the first 3 hours after intubation. Patients were considered to have received a particular therapeutic intervention if the therapy was initiated or administered within 1 hour of diagnosis of cerebral edema. Continuous and ordinal independent variables were first analyzed with logit plots to assess their bivariate association with the outcomes. Transformations were made when appropriate. The logit analysis of hyperventilation and arterial PCO2 versus the log odds of death or a discharge diagnosis of PVS suggested a dichotomous relation best delineated at a PCO2 of 22 mm Hg. We therefore defined hyperventilation as intubation with a mean PCO2 during the first 3 hours after intubation <22 mm Hg. To evaluate the association of clinical outcome with baseline disease severity indicators and therapeutic interventions, we conducted an ordinal logistic regression analysis for the 3-category neurologic outcome variable. Validation was performed by means of bootstrap analysis with 1000 iterations with selection of variables at a P value of .05. Variables were considered significantly associated with the outcome if selected in >50% of the bootstrap ordinal logistic regression iterations.13 All of the statistical computations were 2-tailed and were performed with Stata statistical software (Version 7.0, Stata, College Station, Tex).
RESULTS Of 61 patients identified as having cerebral edema, 13 (21%) died and 4 (7%) survived in a PVS; 36 (59%) patients survived with no persistent neu-
MARCIN ET AL
THE JOURNAL OF PEDIATRICS
VOLUME 141, NUMBER 6 rologic deficits, and 8 (13%) children survived with mild or moderate persistent neurologic disabilities. The age distribution of the children diagnosed with cerebral edema is shown in Fig 1. Three patients were diagnosed with cerebral edema at the time of initial presentation to the emergency department. In the remainder, cerebral edema was diagnosed between 1 and 25 hours after the initiation of therapy for DKA (Fig 2). There was no correlation of clinical outcome with the time of diagnosis of cerebral edema during therapy (Kruskal-Wallis, P = .12). The distribution of the neurologic symptom score and clinical outcome is shown in Fig 3. Of note, all children who either died or survived in a PVS had a neurologic symptom score of 4 or 5 at the time of diagnosis of cerebral edema. Therapies initiated within 1 hour of the diagnosis of cerebral edema are shown in Table II. Mannitol was the most commonly used therapy (n = 41, 67%). Approximately half of the patients were intubated within 1 hour of diagnosis of cerebral edema (n = 33, 54%), but only 17 (28%) had associated hyperventilation (PCO2 <22 mm Hg). Other therapies (glucocorticoids, bicarbonate, and furosemide) were used infrequently and therefore were not included in the multivariate analysis. The results of the ordinal logistic regression are shown in Table III. Variables significantly associated with poor outcomes include a higher neurologic symptom score at the time of diagnosis of cerebral edema, a higher initial BUN concentration, and intubation with associated hyperventilation. In addition, there was a trend toward an association between poor clinical outcome and a lesser rise in measured serum sodium concentration during therapy for DKA. No other variables were significantly associated with clinical outcome. In a subanalysis, we included intubation alone (with or without associated hyperventilation) in place of intubation with hyperventilation in the regression equation. In this subanalysis, intubation alone was
Table II. Therapies initiated within 1 hour of diagnosis of cerebral edema
Outcome
Therapy
Total n = 61
Normal n = 36
Mild-tomoderate disability n=8
Hyperosmolar therapies Mannitol 3% Saline Glucocorticoid therapy Furosemide Bicarbonate Intubation Intubation with hyperventilation
41 (67) 1 (2) 2 (3) 8 (13) 2 (3) 33 (54) 17 (28)
23 (64 1 (3) 1 (3) 1 (3) 1 (3) 13 (36) 4 (11)
5 (63)) 0 (0) 0 (0) 3 (38) 0 (0) 6 (75) 2 (25)
Death or PVS n = 17 13 (76) 0 (0) 1 (6) 4 (24) 1 (6) 14 (82) 11 (65)
Data are given as n (%).
not associated with worse outcomes after adjusting for the other variables (data not shown). The bootstrap analysis was consistent with the ordinal logistic regression analysis. Of the 1000 bootstrap iterations, the same 3 variables (neurologic symptom score, BUN, and intubation with hyperventilation) were selected as significantly associated with poorer clinical outcomes in >50% of the iterations of the logistic regression (Table III).
DISCUSSION In the current study, 3 variables were found to be associated with a poor outcome of children with DKA-related cerebral edema. These variables included an elevated initial BUN concentration, more profound neurologic depression at the time of diagnosis of cerebral edema, and intubation with associated hyperventilation to a PCO2 level <22 mm Hg. Because the neurologic symptom score is a marker for the severity of neurologic abnormalities at the time of diagnosis of cerebral edema, its association with poor outcome is not surprising. The degree of neurologic impairment at the time of diagnosis of cerebral edema has been noted in a pre-
vious study to be correlated with outcome.5 In that study, the author found that children who had had a respiratory arrest at the time of diagnosis of cerebral edema almost uniformly died or survived with severe neurologic deficits.5 The association of elevated initial BUN concentration with poor outcome suggests that patients with more severe dehydration at initial examination are more likely to have adverse outcomes. Previous studies have demonstrated an association between the risk of DKA-related cerebral edema and the degree of initial dehydration.7,14 The findings of the current study therefore strengthen the evidence for a possible pathophysiologic association between dehydration (and possible cerebral ischemia) and cerebral edema in the setting of DKA. Intubation with associated hyperventilation was also associated with a poorer neurologic outcome. This result is consistent with research that has suggested that therapeutic hyperventilation can decrease cerebral blood flow and contribute to cerebral ischemia, potentially exacerbating evolving brain injury. In the setting of traumatic brain injury, the evidence linking hyperventilation and worse neurologic outcomes is growing.15-17 The largest published, 795
MARCIN ET AL
THE JOURNAL OF PEDIATRICS DECEMBER 2002
Fig 1. Age distribution of children diagnosed with DKA-related cerebral edema.
Fig 2. Distribution of clinical outcome and time between initiation of therapy for DKA and diagnosis of cerebral edema.
Table III. Associations of variables with adverse clinical outcome of DKA-related cerebral edema
Coefficient Age (y) –0.084 –0.0091 Initial arterial PCO2 –0.10 Change in serum sodium* Initial BUN concentration 0.086 0.70 Bicarbonate therapy† Neurologic symptom score 2.2 Mannitol therapy –0.053 Intubation and hyperventilation 2.1
95% CI –0.27-0.11 –0.13-0.11 –0.23-0.02 0.01-0.16 –1.13-2.53 1.06-3.37 –1.77-1.67 0.29-3.84
Bootstrap P value analysis (%)‡ .39 .88 .09 .02 .45 .001 .95 .02
7.0 5.1 28.7 59.2 9.6 95.2 9.3 55.3
*Represents change in measured serum sodium concentration during treatment of DKA before diagnosis of cerebral edema. †Administration of bicarbonate before diagnosis of cerebral edema. ‡Percentage of 1000 bootstrap iterations in which variable was selected as significantly associated with outcome.
prospective, randomized trial comparing hyperventilation (PCO2 25 ± 2 mm Hg) versus normoventilation (PCO2 35 ± 2 mm Hg) for acute traumatic brain injury demonstrated that hyperventilation resulted in a significant reduction in the number of patients with a favorable outcome (good or moderately disabled) at 3 and 6 months after injury.18 On the basis of these and other retrospective and animal data, the American Association of Neurological Surgeons has recommended that prophylactic hyperventilation be avoided in the setting of traumatic brain injury.19 Current reviews and texts suggest that in children with DKA-related cerebral edema, prompt control of the airway and ventilation as well as administration of mannitol are the primary therapies,20-22 and glucocorticoid med796
ications are thought not to be effective. These recommendations, however, are based mainly on case reports, because no previous study has investigated the effects of specific therapeutic interventions for this condition.22-24 In the current study, the use of mannitol was not associated with improved outcomes. These findings are similar to those of other studies investigating the use of mannitol in patients with cerebral edema of causes other than DKA, including traumatic brain injury,17,25 stroke,26,27 and hypoxic-ischemic injuries.21,28 In the current study, however, the dosages of mannitol differed among patients (dosages ranged from 0.2 g/kg body wt to 2.0 g/kg). In addition, although all patients categorized as having received mannitol were treated within 1 hour of the diagnosis of cere-
Fig 3. Distribution of clinical outcome and neurologic symptom score. bral edema, the timing of administration in relation to the earliest symptoms of cerebral edema was not known and probably differed substantially among patients. Furthermore, because of the small sample size, our ability to detect some clinically important associations between predictive variables and outcome was potentially limited. It is therefore difficult to draw any firm conclusions regarding the therapeutic efficacy of mannitol from this study. There are several other potential limitations to this study. First, because these analyses were performed on a retrospectively collected data set, the timing of therapeutic interventions and dosages of medications differed among patients. In addition, the administration of therapies was not randomized, so it is likely that therapeutic decisions were biased by disease severity, the time period during which the patient presented, and by physician and institutional preferences. Although we attempted to control for these potential biases by including variables related to disease severity in our analysis and by conducting a multivariate analysis, this bias nonetheless may not have been completely eliminated. In addition, some of the children who met the criteria for cerebral edema did not have the diagnosis confirmed radiographically or by postmortem examination. It is therefore possible that the neurologic abnormalities observed in some of the patients were due to other causes. Furthermore, the severity of neurologic depression at the time of diagnosis, as reflected in the
MARCIN ET AL
THE JOURNAL OF PEDIATRICS
VOLUME 141, NUMBER 6 neurologic symptom score, was determined through review of physician and nursing medical records. These data were recorded in varying degrees of detail and may not have accurately reflected the neurologic state of the patient at the time of diagnosis of cerebral edema. Finally, with regard to the association of intubation with associated hyperventilation and adverse outcome, it is unclear whether this association represents an adverse effect of treatment or possibly reflects the severity of the underlying illness. Although we attempted to control for illness severity and the degree of endogenous hyperventilation at presentation (ie, PCO2), it is nonetheless possible that the observed hyperventilation may reflect either a response to persistent acidosis or neurogenic hyperventilation in response to brain injury. We conclude that children with DKA-related cerebral edema who are most likely to have adverse neurologic outcomes are those who have greater degrees of dehydration as reflected by a higher BUN at the initiation of therapy for DKA and those who have more profound neurologic depression at the time of diagnosis of cerebral edema. Intubation with associated hyperventilation to a PCO2 <22 mm Hg is also correlated with a poorer neurologic outcome, although the current data do not allow us to definitively discern whether this association is an effect of therapy or possibly a reflection of the underlying disease state. Other therapies, including mannitol and intubation in general, were not significantly associated with worse or improved outcomes, although we were limited in our ability to study these associations. The effects of these therapies, in addition to the use of glucocorticoid medications, bicarbonate, and diuretics other than mannitol, warrant further investigation.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
REFERENCES 1. Marcin JP, Slonim AD, Pollack MM, Ruttimann UE. Long-stay patients in
the pediatric intensive care unit. Crit Care Med 2001;29:652-7. Bello F, Sotos J. Cerebral oedema in diabetic ketoacidosis in children. Lancet 1990;336:64. Duck S, Wyatt D. Factors associated with brain herniation in the treatment of diabetic ketoacidosis. J Pediatr 1988;113:10-4. Krane E, Rockoff M, Wallman J, Wolfsdorf J. Subclinical brain swelling in children during treatment of diabetic ketoacidosis. N Engl J Med 1985; 312:1147-51. Rosenbloom A. Intracerebral crises during treatment of diabetic ketoacidosis. Diabetes Care 1990;13: 22-33. Edge JA, Hawkins MM, Winter DL, Dunger DB. The risk and outcome of cerebral oedema developing during diabetic ketoacidosis. Arch Dis Child 2001;85:16-22. Glaser N, Barnett P, McCaslin I, Nelson D, Trainor J, Louie J, et al. Risk factors for cerebral edema in children with diabetic ketoacidosis. N Engl J Med 2001;344:264-9. Harris G, Fiordalisi I, Harris W, Mosovich L, Finberg L. Minimizing the risk of brain herniation during treatment of diabetic ketoacidemia: a retrospective and prospective study. J Pediatr 1990;117:22-31. Mahoney C, Vlcek B, Del Aguila M. Risk factors for developing brain herniation during diabetic ketoacidosis. Pediatr Neurol 1999;21:721-7. Edge JA. Management of diabetic ketoacidosis in childhood. Br J Hosp Med 1996;55:508-12. Multi-Society Task Force on PVS. Medical aspects of the persistent vegetative state (part 1). N Engl J Med 1994;330:1499-508. Multi- Society Task Force on PVS. Medical aspects of the persistent vegetative state (part 2). N Engl J Med 1994;330:1572-9. Chen C, George S. The bootstrap and identification of prognostic factors via Cox’s proportional hazards regression model. Stat Med 1985;4:39-46. Hammond P, Wallis S. Cerebral oedema in diabetic ketoacidosis: still puzzling—and often fatal. BMJ 1992; 305:203-4. Raichle ME, Posner JB, Plum F. Cerebral blood flow during and after hyperventilation. Arch Neurol 1970;23: 394-403.16. C a r m o n a Suazo JA, Maas AI, van den Brink WA, van Santbrink H, Steyerberg
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
EW, Avezaat CJ. CO2 reactivity and brain oxygen pressure monitoring in severe head injury. Crit Care Med 2000;28:3268-74. Schierhout G, Roberts I. Hyperventilation therapy for acute traumatic brain injury. Cochrane Database Syst Rev 2000;2. Muizelaar JP, Marmarou A, Ward JD, Kontos HA, Choi SC, Becker DP, et al. Adverse effects of prolonged hyperventilation in patients with severe head injury: a randomized clinical trial. J Neurosurg 1991;75: 731-9. Brain Trauma Foundation. The American Association of Neurological Surgeons. The Joint Section on Neurotrauma and Critical Care. Hyperventilation. J Neurotrauma 2000;17: 513-20. Rogers MC, Nichols DG, Ackerman AD. Textbook of pediatric intensive care. 3rd ed. Baltimore: Williams & Wilkins; 1996. Mujsce DJ, Towfighi J, Stern D, Vannucci RC. Mannitol therapy in perinatal hypoxic-ischemic brain damage in rats. Stroke 1990;21:1210-4. Roberts M, Slover R, Chase H. Diabetic ketoacidosis with intracerebral complications. Pediatr Diabetes 2001;2: 103-14. Franklin B, Liu J, Ginsberg-Fellner F. Cerebral edema and ophthalmoplegia reversed by mannitol in a new case of insulin-dependent diabetes mellitus. Pediatrics 1982;69:87-90. Shabbir N, Oberfield SE, Corrales R, Kairam R, Levine LS. Recovery from symptomatic brain swelling in diabetic ketoacidosis. Clin Pediatr 1992;31: 570-3. Roberts I, Schierhout G, Alderson P. Absence of evidence for the effectiveness of five interventions routinely used in the intensive care management of severe head injury: a systematic review. J Neurol Neurosurg Psychiatry 1998;65: 729-33. Bereczki D, Liu M, Prado GF, Fekete I. Cochrane report: a systematic review of mannitol therapy for acute ischemic stroke and cerebral parenchymal hemorrhage. Stroke. 2000;31: 2719-22. Bereczki D, Liu M, do Prado GF, Fekete I. Mannitol for acute stroke (Cochrane Review). Cochrane Database Syst Rev 2001;1. Modell JH. Drowning. N Engl J Med 1993;328:253-6. 797