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

Risk Factors for Cerebral Edema in Children with Diabetic Ketoacidosis

Glaser N, Barnett P, McCaslin I, et al.

N Engl J Med 2001;344:264-9. [abstract; full-content for subscribers]

 

Reviewed by Rohit Rao MD and Matthew Scanlon MD, Children's Hospital of Wisconsin

Review posted May 4, 2001

I. What is being studied?:

The study objective:

To identify risk factors for the development of cerebral edema in children with DKA.

The study design:

Multicenter retrospective case control study.

The patients investigated:

All children (persons less/than 18 years of age) in whom cerebral edema related to diabetic ketoacidosis developed between 1982 and 1997 at any of 10 pediatric centers.

II. Are the results of the study valid?

Primary questions:

1.Were there clearly identified comparison groups that were similar with respect to important determinants of outcome, other than the one of interest?

For every child identified with cerebral edema, six controls were chosen. There were two control groups in which the children were placed.

One control group consisted of randomly chosen children with diabetic ketoacidosis. The second group consisted of children with DKA who were matched with the children with the cerebral edema according to age (within 2 years), onset of diabetes (new vs established disease), venous pH at presentation (within 0.1), and serum glucose on presentation within 200 mg per deciliter. If pH values were not available, children were matched according to serum bicarb concentration (within 3 mmol per liter).

2. Were the exposures and outcomes measured in the same way in the groups being compared?

To be included in the study, children identified as having cerebral edema were required to meet all the following criteria: the presence of DKA, an altered mental status and cerebral edema.

The presence of DKA was defined as a serum glucose concentration >300 mg per deciliter [16.7 mmol per liter], a venous pH <7.25 or a serum bicarbonate concentration <15 mole per liter, and the presence of ketones [acetoacetate] in the urine). Alteration in mental status was defined as obtundation or disorientation. The criteria used for cerebral edema included children with either radiographically or pathologically confirmed cerebral edema, or specific treatment for cerebral edema (hyperosmolar therapy or controlled hyperventilation) that was followed by clinical improvement.

Because cerebral infarction may occur as a consequence of cerebral edema or impending herniation, a neuropathologist also reviewed radiographic studies of children with cerebral infarction. Six patients whose radiographs revealed patterns of infarction consistent with the consequences of cerebral edema were included in the cerebral-edema group.

The records of all children who died (with or without type 1 diabetes) at those 10 centers during the study period were reviewed to ensure that no cases were missed. The retrospective review may have led to an underestimation of the number of cases of cerebral edema because edema can occur subclinically. Additionally, the inclusion of children with a perceived clinical response to interventions made to treat cerebral edema may have led to misclassification of some children who did not truly have cerebral edema.

The classification of degree of acidosis was also problematic. One problem is that all patients did not have arterial blood gas analysis performed. Consequently, venous pH and PCO2 was converted to arterial values using mathematical formulas. This is not standard, accepted practice in most ICUs and may have led to some inaccuracy when assigning any given patient a degree of acidosis. One solution may have been the presentation of base deficits as another indicator of metabolic acidosis.

One investigator at each center recorded all data. To assess interrater agreement in the recording of data, 10 percent of the records were randomly selected and examined by a single investigator. Kappa of 0.9 reflected a good interrater agreement. Having said this, the authors did not provide enough information for the readers to know whether the Kappa statistic is relevant. As only one investigator recorded data at a given institution; within that institution, there is no need for "interrater" reliability. As the Kappa is usually used in evaluating agreement between two people (classically, two people independently reviewing the same data), it may have no value in comparing whether two investigators at separate institutions were interpreting separate data in the same manner.

The same statistical tools were used for examining data from all groups.

3. Was follow-up sufficiently long and complete?

Yes. The end point was development of cerebral edema. So the follow up was complete. 12% of data points were missing and they were imputed but that data is unavailable to the reader.

Secondary questions:

4. Is the temporal relationship correct?

Cerebral edema is a known complication of diabetic ketoacidosis that can develop with a variable delay after presentation. The authors present data on the timing of cerebral edema after initiation of therapy. Some patients presented initially with cerebral edema, while the vast majority of patients were diagnosed at between 3 and 13 hours .

5. Is there a dose-response gradient?

Yes, logistic regression analysis demonstrated an inverse relationship between pCO2 on admission and a direct relationship with BUN.

III. What are the results?

1. How strong is the association between exposure and outcome? How precise is the estimate of the risk?

The authors estimated the relative risk of the different variables. Using the case-control data, the authors computed the odds ratio and used that as an estimate of the relative risk. This is a reasonable procedure when the disease is rare in the population. The relative risk cannot be directly determined in case control studies because the investigators choose both the number of cases and controls and the resulting proportions.

Odds ratios were considered significant if their confidence intervals did not cross 1.

Comparing the cases to the matched controls, these included:

• Initial BUN (per increase of 9 mg/dl): 1.8 (95% CI: 1.2-2.7); p value 0.008
• Initial partial pressure of arterial carbon dioxide (per decrease of 7.8 mm Hg): 2.7 (95% CI: 1.4-5.1); p value 0.002
• Treatment with bicarbonate: 4.2 (95% CI 1.5-12.1); p value 0.008
• Rate of increase of serum sodium concentration (per increase of 5.8 mmol/l/hr):0.6 (95% CI: 0.4 - 0.9); p value 0.01

Using the lower limit of the 95% confidence intervals, the odds of cerebral edema almost double for every 9 mg/dl increase in BUN (average in controls ~20 mg/dl) or decrease in pCO2 by ~8 mm Hg (controls' average ~18 mm Hg), while the odds quadruple with bicarbonate treatment. In addition, the slower the rise in serum sodium during treatment, the more likely cerebral edema is to develop.

Another illuminating aspect of this study is that it suggests that rates of fluid administration or correction of serum glucose do NOT add to the risk of cerebral edema. This has been clinical dogma passed down over the years and may be more securely laid to rest by this study. However, it is unclear if all patients were given fluids similarly and glucose reduced at the same rate in which case, this assumption might not be accurate.

Since the incidence of cerebral edema is small in DKA (~1% in this study), a relative risk of 2 would roughly translate to a 2% risk of cerebral edema, still rather small.

One factor that may influence the precision of the risk estimate is the issue of missing 12% of continuous data. The implication of missing 12% of data is significant. The authors did not provide information related to either the type of information that was missing nor the nature of how that data was missing. While imputation is an accepted statistical tool for dealing with missing data, the validity of this technique depends on whether date is missing at random (MAR) or missing completely at random (MCAR). Without such information, the reader is left to question this design methodology. Misuse of imputation of the missing data could seriously alter the risk estimated.

The potential misclassification of patients into the cerebral edema or no cerebral edema category could also alter the precision of the risk estimate. As discussed in section II 2 above, the clinical criteria used for classifying cerebral edema increase the chance of inappropriately attributing risk.

IV. What are the implications for my practice?

1. Are the results applicable to my practice?

As a pediatric intensivist, the patients in this study reflected the patient population I care for.

2. What is the magnitude of the risk?

The risk of cerebral edema in patients with DKA ranges from 1.2 to 1.5, depending on the risk factor. An elevated BUN on presentation has the least impact on risk. The risk increases with a low PCO2 on presentation and with the use of bicarbonate infusions, respectively.

This study makes it quite clear why it has been so difficult to reliably identify the risk factors for CE in DKA patients. Because the baseline risk is so low, a large number of patients would need to be studied prospectively to prove that one therapy or risk factor altered the patients' course. The case control study methodology, identifying a suitably large number of cases, is ideally suited to identification of the risks.

3. Should I attempt to stop the exposure?

Unfortunately, two of the risk factors are laboratory findings at the time of presentation. It is not realistic to think that these factors might be influenced beyond efforts to improve education of existing diabetics to minimize the frequency of episodes of DKA. However, we can use this information to be at heightened awareness for the possible development of CE in patients with these added risk factors.

The one factor that pediatricians can directly influence is the use of bicarbonate for DKA. Bicarbonate is used to treat metabolic acidosis. This ignores the issue that the correct treatment for ketoacidosis is insulin. The bicarbonate use merely masks the ketoacidosis. Recognizing that the bicarbonate is not necessary in the treatment of DKA is a first step. Furthermore, this study suggests that bicarbonate use increases the risk of cerebral edema. The lack of indication for using bicarbonate coupled with the potential risk supports the ending of the use of bicarbonate in the treatment of DKA in pediatric patients.

Comments

Back to the EB Journal Club Index