Criteria abstracted from The Users' Guides to the Medical Literature series in JAMA
Glucocorticoid therapy for hypotension in the cardiac intensive care unit.
Millar KJ, Thiagarajan RR, Laussen PC.
Pediatr Cardiol 2007 28(3):176-82.[abstract]Reviewed By: RE Kirsch MD, L Zorzela MD, D Garros MD, University of Alberta, Stollery Children
Review posted January 15, 2009
I. What is being studied?:
Glucocorticoid therapy for hypotension in children with congenital cardiac disease.
The study objective:
â€ś To describe the effects of glucocorticoid treatment on hemodynamics and mortalityâ€ť in patients in a pediatric cardiac intensive care unit and â€śto identify in which patients glucocorticoids may be useful.â€ť
The study design:
Retrospective chart review between July 1, 2001 and June 30, 2003 of children younger than 16 years with cardiac disease treated with glucocorticoids for hypotension.
II. Are the results in the study valid?
Was a representative group of patients completely followed up?
Yes, the population studied is representative of a mixed medical-surgical pediatric cardiac intensive care unit (CICU). Patients reviewed were younger than 16 years of age with cardiac disease treated with glucocorticoids for hypotension in a CICU across a 2 year time period. Of patients receiving glucocorticoids, most were infants (88.2% younger than 1 year old), although age ranged from 12 days to 11 yrs 9mos. Most study patients meeting inclusion criteria had structural congenital heart disease (92%), the majority of whom had steroids given after palliative surgery for congenital heart disease (65%). The most common palliation was Stage 1 repair for HLHS (11 patients), although 9 had right-sided obstructive lesions, 9 had left-sided obstructive lesions, and 7 had VSD closure. 6 patients had ECMO during the admission examined. Those patients who had received enteral, parenteral, or inhaled glucocorticoids in the month prior to CICU admission were excluded. The patients were identified from the pharmacy list of CICU patients who had received IV glucocorticoids between July 1, 2001 and June 30, 2003 to ensure all potential candidates for study were identified. Glucocorticoids were prescribed for refractory hypotension. Hypotension was considered refractory after: adequate preload was ensured (10mL/kg given to target an atrial pressure dependant on monitoring and patient anatomy), ionized calcium was appropriate (1.2-1.3 mmol/L), Dopamine (5-10mcg/kg/min) or Epinephrine (0.05 mcg/kg/min) was titrated for target blood pressure, contributing factors (ex. fever, anemia, effusion, residual defects) had been evaluated and treated. Steroid was generally given after drawing a baseline serum cortisol level but without waiting for the test result to treat and was given in conjunction with prophylactic IV ranitidine and oral nystatin.
Was follow-up sufficiently long and complete?
Yes and no. The investigators defined a favorable hemodynamic response to steroids as an increase in mean blood pressure of at least 20% without an increase in inotrope score during the 24 hours following steroid initiation. They collected data on inotrope usage, blood pressure, and other surrogate markers of cardiac output from 12 hours prior to steroid therapy until 24 hrs following steroid therapy. All patients had 12 hours of pretreatment data recorded. 2 patients did not have 24 hrs of data following first administration of steroids because one went for an arterial switch operation 18 hours after steroid was given, and one was placed on ECMO at 9 hours post steroid administration. 24 hrs is an appropriate time period to find a hemodynamic response to steroids. However, patients were not followed for a set period of time, such as â€ś30 daysâ€ť for outcomes such as mortality, duration of hospital stay, duration of CICU stay, duration of mechanical ventilation, or neurologic outcomes. Hence no long term outcomes can be correlated. It is therefore not clear, for example, whether the mortality they report is CICU or hospital mortality.
Were all potential predictors included?
Yes and no. Patient demographics, diagnosis, indications for therapy, course in ICU, complications and outcome were recorded. Appropriate logistic regression analysis examined potential predictors of glucocorticoid response. The secondary outcome of mortality was also analyzed in relationship to the chosen clinical parameters, including the clinical response to steroids.
There is no clear evidence in the literature as to what may predict a hemodynamic response to glucocorticoids, but reasonable parameters were certainly examined. Patient age, blood pressure, inotrope score, serum cortisol, previous ECMO, and previous surgery were examined in univariate analysis for predictors of hemodynamic response. However, when predicting a hemodynamic response to steroids, one wonders if the response will be greater if the hemodynamics at baseline were poorer (i.e., the presence of a clinically more severe low cardiac output syndrome would predict a more dramatic response to steroids). To this end, it would be informative to know other indicators of low cardiac output and attempt to see if they predicted a positive hemodynamic response to steroids. Some of these predictors would be cardiopulmonary bypass time, cross-clamp times, the serum lactate pre- or post-operatively, and the number of days of mechanical ventilation pre-operatively (1).
Predictor variables were also examined for mortality. These included age, serum cortisol, surgery, blood pressure, inotrope score before steroids, hemodynamic response to steroids and sepsis.
Did the investigators test the independent contribution of each predictor variable?
Logistic regression for response to steroids and mortality was performed using the stated clinical parameters.
Were outcome variables clearly and objectively defined?
Yes. The values used to judge a blood pressure change were defined at the outset, the inotrope score was plainly outlined, the criteria for infection (positive culture resulting in beginning or changing antimicrobial therapy), and for GI bleed (blood loss requiring transfusion) were explicitly defined.
III. What are the results?
What is(are) the prediction tool(s)?
There is no prediction tool found in this study.
There was a subset of patients that had a significant improvement of hemodynamics (increase in mean blood pressure of â‰Ą 20% without increase in inotrope score) following steroid administration (21 patients, 41%). Each parameter of hemodynamic response showed a statistically significant improvement with tight confidence intervals at 24 hrs post steroid therapy. Mean blood pressure was 7.8 mmHg higher (95% CI 5.0-10.6), heart rate was 11.8 min -1 lower (95% CI 6.2-17.5), volume infused was 1.4 mL/kg/hr less (95% CI 0.8-2.1), urine output increased (p=0.001), and inotrope score declined (p<0.0001) compared with baseline at 12 hrs prior to starting steroid. However, none of these parameters, the patient demographic data, or diagnostic data predicted a favorable response to steroids. Nor did the patient random serum cortisol level [20 patients (46.5%) had a serum cortisol <5mcg/dL]. The lack of predictability for hemodynamic response to steroids was also observed in the Corticus study. Here, in adult patients, corticotropin responsiveness did not have an association with shock reversal (2).
Interestingly, all of the 6 patients who died showed no hemodynamic response, and the analysis was able to predict that having a hemodynamic response to steroids predicted survival. All 21 â€śrespondersâ€ť survived, while only 80% of the â€śnon-respondersâ€ť lived. This represents an OR of death of 11.4 (95% CI 0.6 â€“ 214.6) for the non-responders (p = 0.036 by Fisher's exact test). The only other predictor of death was the presence of sepsis (OR 8; 95% CI 1.3-50.8), though it is unclear if this was a univariate association they made, or part of the multiple logistic regression, whereby the result would be independent of steroid responsiveness.
What the study shows, then, is a statistically significant association between a lack of steroid responsiveness and the likelihood of patient death. It also suggests the patient is more likely to die if they have sepsis. Unfortunately, this doesn't clarify the muddy waters regarding indications for steroid administration and may only be demonstrating what already seems clinically logical â€“ if a patient has refractory hypotension that stays refractory, they are very likely to die.
How well does the model categorize patients into different levels of risk?
There is no categorization into levels of risk because no potential predictors studied proved to be predictive of response to steroids. It seems only that there is improved survival IF a positive hemodynamic response occurs, but it remains unknown what predicts hemodynamic response.
How confident are you in the estimates of risk?
One can be confident that the response found was precise in the tight confidence intervals. However, there can be no estimate of risk (or predictor of response) to steroids because no variable studied correlated with response. With regards to mortality risk, the confidence intervals around the OR of 11.4 for non-responders are wide ( 0.6 â€“ 214.6 ). With regards to sepsis as a predictor of mortality, the confidence intervals are very wide (OR 8, CI 1.3-50.8) as one would expect with the small sample size involved (51 patients met inclusion criteria).
IV. Will the results help me in caring for my patients?
Does the tool maintain its prediction power in a new sample of patients?
There is no prediction tool available from this study and no other patients were studied.
Are your patients similar to those patients used in deriving and validating the tool(s)?
The study patients are similar to those admitted in our institution as a whole (a moderate volume congenital heart disease program). However, we do not admit pre-operative cardiac neonates to the PICU so the study group varies from our practice in that manner. There is, however, a similar timeline to initiation of steroid therapy in our PICU as to that outlined in the paper.
Will the results lead directly to selecting or avoiding therapy?
These results support that some patients respond to steroid therapy and that this response involves improved hemodynamic parameters and a decrease in inotrope requirement. However, it does not provide any ability to select patients that might respond to steroids, nor is it able to prove an outcome benefit beyond 24 hour hemodynamic parameter improvements, as the mortality data, and data on infection rates in steroid recipients represent very small numbers. It remains unclear as to when to select therapy (or avoid therapy) and in whom, and remains in the realm of clinical judgment on a patient by patient basis. The risk of using the therapy (steroids) with the ongoing debate of potentially increasing nosocomial infection rates remains undefined.
Are the results useful for reassuring or counseling patients?
Not particularly. It does lend support to the anecdotal evidence of decreased inotrope requirements and improved hemodynamics after steroid administration in a series of CICU patients. It continues to be hoped, but remains unproven that a short term hemodynamic benefit translates into long term outcome benefits without excess risk. To answer this question, an RCT would be required to truly delineate the benefits and risks of steroid therapy separate from other interventions required for these complex, and high acuity patients.
- Cheung PY, Chui N, Joffe AR, Rebeyka IM, Robertson CM. Western Canadian Complex Pediatric Therapies Project, Follow-up Group. Postoperative lactate concentrations predict the outcome of infants aged 6 weeks or less after intracardiac surgery: A cohort follow-up to 18 months. J Thorac Cardiovasc Surg. 2005 Sep;130(3):837-43.
- Sprung CL, Annane D, Keh D, Moreno R, et al. Hydrocortisone Therapy for Patients with Septic Shock. N Engl J Med. 2008 Jan; 358:111-24.
Last Updated: January 15, 2009