Criteria abstracted from The Users' Guide to Medical Literature, from the Health Information Research Unit and Clinical Epidemiology and Biostatistics, McMaster University

Highlighted lines and questions below provide links to the pertinent description of criteria in The EBM User's Guide, now available at the Canadian Centres for Health Evidence

Article Reviewed:

Please visit the new Evidence Based Journal Club Reviews

Factors associated with adverse outcomes in children with diabetic ketoacidosis-related cerebral edema.

Marcin JP, Glaser N, Barnett P, et al.

J Pediatr. 2002;141(6):793-7. [abstract]

Reviewed by Curt Kennedy, MD, Texas Children's Hospital

Review posted July 7, 2003

I. What is being studied?

Study objective:

The objective of this study was to determine which factors, if any, were associated with adverse outcomes in children diagnosed with DKA related cerebral edema. The factors included in the analysis included biochemical abnormalities, neurological symptoms and signs at the time of diagnosis of cerebral edema, and therapies used to manage cerebral edema.

Study design

This is a multi-institutional retrospective cohort analysis (risk factor = DKA related cerebral edema). Data was from 10 pediatric centers, 1982-1997, from all patients 18 years or less with a diagnosis of DKA related cerebral edema.

Inclusion criteria

Inclusion criteria included:

1. DKA (glucose > 300 mg/dL, venous pH < 7.25 OR serum bicarb < 15 mmol/L, ketonuria)
2. Altered Mental Status (obtundation or disorientation)
3. Additional confirmation
   1. Radiographic / pathologic confirmed cerebral edema
   2. Or - treatment for cerebral edema followed by improvement

Outcomes evaluated:

Outcome studied was discharge status:

1. Death or PVS (category of interest)
2. Survival with mild to moderate neurologic disability
3. Survival with a normal neurologic status

Analysis type

Multiple logistic regression with bootstrap validation with two groups of independent variables: 1) risk factors for DKA related cerebral edema and 2) interventions directed towards treating cerebral edema.

II. Are the results of the study valid?

Note: These questions follow from Randolph AG et al. Understanding articles describing clinical prediction tools. Crit Care Med 1998;26:1603-1612. [abstract]

1. Was a representative group of patients completely followed up?

The patient selection process was well designed by the apriori definition of DKA and cerebral edema. Bias was minimized by the fact that both physicians were blinded during patient outcome categorization.

The weakness of the study lies in the diagnosis of cerebral edema, in the clause "clinical improvement after a specific treatment for cerebral edema." (p. 794, Methods Ð column 1). This statement allows for the possibility to include people that really do not have DKA related cerebral edema, since patients can improve for a variety of reasons that may not necessarily be related to the treatment provided.

Also, they did not provide information about the group of patients other than age. No info on demographics, new diabetics vs established diabetics, socio-economic status, etc. A good study would have included a diverse population of children.

2. Was follow-up sufficiently long and complete?

Categorization was based upon discharge status. No additional long-term information was available.

3. Were all potential predictors included?

The risk factors for developing DKA related cerebral edema were established in a prior study , referenced in this article. The assumption is that the referenced study included all potential predictors. The addition of therapeutic interventions provided for a more comprehensive set of predictor variables.

A relevant predictor that we feel should have been included is the risk of death based on year of treatment, since practice patterns surely changed over the 15 years studied.

4. Did the investigators test the independent contribution of each predictor variable?

Univariate analysis results were neither listed in the body of the results nor in a separate table. However, the raw data is presented in Table II and statistical figures can be calculated by setting up 2x2 tables. These tables are listed below, along with the associated p-value for each variable. (Note: Total N = 61, Total Death/PVS = 17. Tables that contain < 5 cases in more than one cell violate assumptions of the 2x2 table and are therefore not included. These variables are: 3% saline, Glucocorticoid, Furosemide and Bicarbonate therapies.)

Mannitol: p = 0.51

	Death/PVS	Norm/Mild	Total
Mannitol	13	28	41
No Mannitol	4	16	20
Total	17	44	61

Intubation: p=0.01

	Death/PVS	Norm/Mild	Total
Intubation	14	19	33
No Intubation	3	25	28
Total	17	44	61

Intubation w/Hyperventilation (I w/ H): p<0.001

	Death/PVS	Norm/Mild	Total
I w/ H	11	6	17
No I w/ H	6	38	44
Total	17	44	61

Logistic regression analysis was conducted including all variables and coefficients of the regression equation are presented.

5. Were outcome variables clearly and objectively defined?

Patients were classified into three clinically important categories: Death/PVS vs. mild-moderate disability vs. normal.

Specific criteria for distinguishing persistent vegetative state from a moderate/severe disability are not specifically stated. An established scale, such as the Pediatric Cerebral Performance Category (PCPC) would have been a more objective measure.

However, the fact that two blinded physicians agreed on all but two patients helps reassure the validity of the study.

III. What are the results?

1. What is(are) the prediction tool(s)?

The coefficients for the regression equation are given in Table III along with the 95% CI, the p value and the percentage of times the variable was included in the bootstrap analysis. Significant predictors, coefficients (95% CI for coefficients) and p-values for the logistic regression equation were: 1) Initial BUN concentration = 0.086 (0.01-0.16): p=0.02; 2) Neurologic Symptom Score = 2.2 (1.06-3.37): p<0.001; 3) Intubation and Hyperventilation = 2.1 (0.29-3.84): p = 0.02. Initial arterial PCO2, Change in serum sodium, Bicarbonate therapy and Mannitol therapy did not reach statistical significance.

The correction coefficient is not included, so the equation is incomplete. However, the purpose of the paper was to identify risk factors, not create a predictive equation.

2. How well does the model categorize patients into different levels of risk?

Again, a predictive model is not presented. Risk factors of

1. Higher neuro symptom score at time of Dx. of cerebral edema
2. Higher BUN
3. Intubation with hyperventilation

are presented as significantly associated with adverse outcomes.

3. How confident are you in the estimates of the risk?

The estimates of risk are not explicitly stated, except as a coefficient in the multivariate analysis (Table III). [Coefficients can be used to calculate an odds ratio, where e^{(coefficient)} = odds ratio.]

Therefore, for Initial BUN concentration, the OR = e^(0.086) = 1.09 and the 95% CI = e^(.01) to e^(.16) = 1.01 to 1.17. In other words, for each unit increase in BUN, the risk of adverse outcome is 1.09 times higher.

For Initial Neurologic symptom score, the OR = e^(2.2) = 9 and the 95% CI = e^(1.06) to e^(3.37) = 2.9 to 29. Therefore, the risk of an adverse outcome increases nine-fold for each unit increase in the neurologic score.

For Intubation and hyperventilation, the OR = e^(2.1) = 8.2 and the 95% CI = e^(0.29) to e^(3.84) = 1.3 to 46.5. This is a dichotomous variable; one was considered either intubated and hyperventilated or not. Therefore, the risk of an adverse outcome increases eight-fold if intubation and hyperventilation occur.

If death and PVS are separated as one outcome and normal and mild to moderate disability are separated as the other, then a 2x2 table can be generated from the data in Table II to calculate an odds ratio for intubation and hyperventilation. This 2x2 table is noted above: the associated Odds Ratio (OR) = Odds of exposure in death / Odds of exposure in controls = [11/6] / [6/38] = [11x38] / [6x6]= 11.6 [95% CI = 3.1 to 43.2]

(NOTE: It should be noted in this small series, it may be very difficult to confidently isolate treatments from underlying condition. In particular, it may be that only the most severely compromised patients were intubated and hyperventilated, and that it is their severity of cerebral injury on presentation that is the true etiology of their outcome, rather than the therapy employed. The authors readily acknowledge this concern.)

IV. Will the results help me in caring for my patients?

1. Does the tool maintain its prediction power in a new sample of patients?

The validation was based on a bootstrapping technique. There was not an independent sample of patients against which the findings were validated. This is, of course, not surprising given the relative rarity of cerebral edema in DKA. Bootstrapping validation is an acceptable method of internal validation on small data sets, but it should be noted that two of the predictors (BUN and Intubation w/ hyperventilation) were present in < 60% of the bootstrap models.

2. Are your patients similar to those patients used in deriving and validating the tool(s)?

Patient populations were similar at baseline. (They share the same demographics, same disease, same presentation and complications). However, they differ with respect to therapeutic interventions. (Practice styles differ now than they did in 1982.)

3. Will the results lead directly to selecting or avoiding therapy?

Hyperventilation has been shown to adversely impact outcome in a number of clinical conditions, most notably in traumatic brain injury. We avoid maintaining pCO2 < 35 in any population, so this article supports that practice.

Are the results useful for reassuring or counseling patients?

The article identified 3 factors as significantly associated with adverse outcomes: Initial BUN, Initial Neurological Symptom Score, and Intubation with hyperventilation.

1. The association with BUN is quite weak and of little use.
2. Initial Neurological Symptom Score had a strong association with a clinically significant Odds Ratio (At least 3X likely to die or have PVS). Also, approximately 70-80% of the patients that had the most severe presentation (Levels 4 and 5 on Figure 3) had CNS dysfunction upon discharge or expired. This information may have utility in selected circumstances.
3. Intubation with hyperventilation: was discussed above.

References:

1. Glaser N, Barnett P, McCaslin I, et al. Risk factors for cerebral edema in children with diabetic ketoacidosis. N Engl J Med 2001; 344:264-9. [abstract]; PedsCCM EB Journal Club Review
2. The Brain Trauma Foundation. The American Association of Neurological Surgeons. The Joint Section on Neurotrauma and Critical Care. Hyperventilation. J Neurotrauma 2000 Jun-Jul;17(6-7):513-20. [abstract]

 

Comments

Back to the EB Journal Club Index