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

Association of timing, duration, and intensity of hyperglycemia with intensive care unit mortality in critically ill children.

Srinivasan V, Spinella PC, Drott HR, Roth CL, Helfaer MA, Nadkarni V.

Pediatr Crit Care Med. 2004 Jul;5(4):329-36. [abstract]

Reviewed by Andora Bass MD, Duke University Medical Center, Durham, NC

Review posted March 1, 2005

I. What is being studied?

Study objective:

There were two objectives of this study: 1) to determine the effects of hyperglycemia on mortality in critically ill children and 2) to assess the feasibility of a prospective randomized controlled trial on strict glucose control in this population.

Study design

The study is designed as a retrospective review of consecutive PICU admissions during a six month period.

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?

Yes. The authors retrospectively reviewed 152 consecutive critically ill children between 1 month and 21 years of age admitted to the PICU in a six month period. Critically ill was defined as the need for mechanical ventilation or vasoactive infusions (except dopamine ≤ to 3 mcg/kg/min) for more than 24 hours. They excluded postoperative cardiac and solid organ transplant patients. They also excluded patients with insulin-dependent diabetes mellitus and those that died within 24 hours of ICU admission.

Was follow-up sufficiently long and complete?

The follow-up was complete in that it assesed the outcome of mortality in their population. The patients were followed through their entire PICU admission but no follow-up beyond their hospitalization was attempted.

2. Were all potential predictors included?

Many potential predictors of mortality or hyperglycemia-related mortality were included in statistical analysis:

• blood glucose level at 12 and 24 hours
• peak blood glucose level
• median time to peak blood glucose in hours
• duration of hyperglycemia (>126 mg/dL) in percent PICU days
• median blood glucose levels in first 24 hours, in first 48 hours, and in entire PICU stay
• median frequency of blood glucose sampling, in times per day

The authors also looked at age, mechanical ventilation, vasoactive infusions, epinephrine, steroids, steroid dose, 12 hour PRISM score, and 24 hour PRISM score. Factors which were not evaluated by the authors but which could potentially be prognostic might include presence of seizure (and/or administration of phenytoin), and presence of documented infection.

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

Each predictor variable had an odds ratio calculated by univariate analysis and multivariable logistic regression was performed. However, many of the glucose variables cannot be strictly considered to be independent and no mention is made in the text to account for multicollinearity.

4. Were outcome variables clearly and objectively defined?

The outcome assessed by this study was clear: mortality during PICU course. (Mortality in the study population was 23/152, or 15%.)

III. What are the results?

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

The authors do not attempt to formulate a prediction tool to classify the study patients into specific levels of mortality risk.

However, what they did find was that patients with higher peak blood glucose levels, higher % of PICU days with hyperglycemia, and median blood glucose levels > 150 for the first 48 hours had a higher risk of mortality. Also, there was an 8.3% mortality rate among patients with a 24 hour blood glucose < 100 while the mortality rate rose to 28.5% in those with a 24 hour blood glucose > 180.

In multivariate analysis, the authors found four significant variables that were independently predictive of mortality: 24-hr PRISM score, use of epinephrine, peak blood glucose, and duration of hyperglycemia. When adjusted for age, severity of illness, and use of epinephrine, two variables remained independently associated with mortality in this population: peak blood glucose and duration of hyperglycemia.

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

Although the authors calculate odds ratios to define the magnitude of association between the predictor variables and ICU mortality in the study population, they do not provide a means to prospectively differentiate between patients with higher vs. lower risks of mortality. By univariate logistic regression analysis the peak blood glucose and the intensity and duration of hyperglycemia during the PICU stay defined the risk. Risk was confirmed using multiple logistic regression and again these two variables were predictive of mortality.

Prognostic Variable	Odds Ratio of death	95% CI	p value
BS > 150 at 24 hours	3.4	1.4 - 8.6	< 0.01
Hyperglycemia (> 126) for > 50% of PICU days	5.9	2.4 - 14.7	< 0.001
Median BS > 150 for first 48 hours	2.96	1.06 - 8.33	< 0.05

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

The confidence intervals above are rather wide, but this is to be expected in a relatively small study. It is important to note that in the multivariate analysis, the adjusted odds ratios appear close to one (e.g., duration of hyperglycemia's adjusted OR is 1.03, 95% CI 1.01 - 1.05) because duration is a continuous variable. In other words, for each hour of hyperglycemia, the risk of mortality increases by 1.03 times.

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 authors do not evaluate a new sample of patients. We cannot be sure that the predictor variables evaluated in this study will be associated with mortality to a similar degree in another cohort of critically ill children.

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

The patients in this study are similar to most non-cardiac PICU's. However, based on their Table 1, we do not know if non-trauma surgical patients or if bone marrow transplant patients were included. This difference could have an impact because of the steroids used in these populations (needed for immunosuppression and post-operative management after neurosurgical resections and airway surgeries). It is likely that hyperglycemia in these patients would be more common but it is unknown if it would lead to higher mortality. Therefore, it would be difficult to apply these results in a PICU that had a significant number of surgical or bone marrow transplant patients. It would also be difficult to apply these results in a PICU with a mortality rate lower than that of the unit in the study (15%).

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

This study will make pediatric intensivists more aware of the role of hyperglycemia in mortality. However, the results of this study warrant cautious consideration because they are based on only 23 outcome events (mortality) accrued from a heterogeneous population in a single center. Tight glucose control with insulin is more likely to be discussed and probably utilized than in the past, but additional study will be needed in order to determine which patients are likely to benefit from it.

4. Are the results useful for reassuring or counseling patients?

Yes. This study reports a significant association between peak blood glucose levels and duration of hyperglycemia in critically ill children and mortality. Having this data will make it easier to explain to patients and their parents the need for strict glucose control. Many people are hesitant to add new therapies to the treatment plan without evidence of the potential benefits. It will also reassure them that the risk of hypoglycemia in this study was not more likely to occur in the patients who were treated with insulin infusions, although this conclusion is based on a small number of patients. A prospective randomized controlled trial is warranted in order to determine the risks and benefits of instituting aggressive blood glucose control in critically ill children.

References

1. Capes SE, et al: Stress hyperglycemia and prognosis of stroke in nondiabetic and diabetic patients: A systematic overview. Stroke 2001; 32:2426-2432. [abstract]
2. Capes SE, et al: Stress hyperglycemia and increased risk of death after myocardial infarction in patients with and without diabetes: A systematic overview. Lancet 2000; 355:773-778. [abstract]

 

Comments

Back to the EB Journal Club Index