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

Outcome of children requiring admission to an intensive care unit after bone marrow transplantation.

Jacobe SJ, Hassan A, Veys P, Mok Q.

Crit Care Med. 2003;31(5):1299-305. [abstract]

Reviewed by Eric Harry MD, Children's Hospital and Regional Medical Center, Seattle, WA

Review posted February 4, 2004

I. What is being studied?:

The study objective:

To review the outcome of bone marrow transplant recipients admitted to a pediatric intensive care unit and attempt to identify admission characteristics that might accurately predict a poor outcome.

The study design:

Retrospective case-note review.

The outcomes assessed:

Three measured outcomes were included in the study: 1) survival to discharge from the ICU, 2) survival for 30 days after ICU discharge and 3) survival to 6 months.

II. Are the results in the study valid?

Primary questions:

1. Was there a representative and well-defined sample of patients at a similar point in the course of the disease?

Yes. 210 bone marrow transplants were performed at Great Ormond Street Hospital for Children in London during the 5-year study period from 1994-1998. Forty of these 210 patients accounted for 57 PICU admissions during the study period. Readmissions were treated as separate patient encounters and analyzed accordingly. The forty patients included in the analysis had a wide variety of underlying diagnoses leading to BMT, with the most common being immunodeficiency (n=16) and hematological malignancy (n=14). The wide variety of underlying diagnoses and limited numbers of patients for each diagnosis limits the generalizability of these results.

2. Was follow-up sufficiently long and complete?

Yes. Follow-up to death or survival at 6 months was determined in all 40 patients or all 57 admissions. The median follow-up of survivors was 587 days (absolute range 308-1803 days). Six patients were followed up differently: 4 patients discharged from the ICU for palliative care died within 1 week of ICU discharge and were classified as not surviving to ICU discharge, while 2 patients discharged from the ICU for palliative care survived longer than 1 week after ICU discharge and were counted as ICU survivors.

Secondary questions:

3. Were objective and unbiased outcome criteria used?

Yes. The three primary outcomes were all mortality based (survival to ICU discharge, 30 days after ICU discharge and 6 months) and were thus relatively objective and unbiased.

4. Was there adjustment for important prognostic factors?

Yes. Risk factors analyzed in the study included preadmission characteristics (underlying diagnosis, BMT type, conditioning regimen, CMV status, significant Graft vs. Host Disease), Pediatric Risk of Mortality (PRISM) score, Pediatric Index of Mortality (PIM) score, multiple organ system failure, total serum bilirubin, need for mechanical ventilation (including oxygenation index and need for HFOV), administration of inhaled nitric oxide or surfactant, inotrope use (defined as either dopamine or dobutamine > 5 mcg/kg/min or any dose of epinephrine or norepinephrine), total parenteral nutrition use and renal replacement therapy (continuous venovenous hemofiltration).

It is not clear why they were unable to make adjustments for important prognostic factors, i.e., they only report univariate analysis of factors. For instance, they could state that the duration of intubation was shorter in survivors than non-survivors, and that the PRISM scores in survivors were less than those of non-survivors, but the reported mortality numbers were not adjusted and no multivariable modeling was performed. Even though they did find statistically significant differences in MOSF scores and HFOV use, they did not factor them into their mortality estimates. Finally, the most common mode of death in the study was withdrawal of support because of "no hope," a very difficult mode of death to factor in but one that is clearly not objective.

III. What are the results?

1. How large is the likelihood of the outcome event(s) in a specified period of time?

Overall survival for BMT recipients admitted to the ICU was poor, with an overall 6-month mortality of 72.5% (11 patients of the 40 patients/57 admissions survived to 6 months), though 32 of 57 admissions (56.1%) survived to ICU discharge. 21 of these 32 survivors were alive at 30 days and 11 survived to 6-month follow-up. None of the preadmission characteristics predicted survival, and both the PRISM and PIM calculated risk of mortality underestimated the actual mortality of this patient group.

The three main outcomes were:

1. survival to discharge from the ICU: 32/57 admissions or 56.1%
2. survival for 30 days after ICU discharge: 21/57 admissions or 36.8%
3. survival to 6 months: 11/57 admissions or 19.3% (11/40 patients or 27.5%)

Additionally, the authors describe the characteristics of some sub-populations: Mortality for patients requiring intubation was 87%, and there were no survivors among patients with respiratory failure secondary to a pulmonary infection requiring mechanical ventilation > 48 hours. There were also no survivors among patients requiring HFOV (10 patients), inhaled NO (4 patients) or CVVH (3 patients). Survival was also related to the number of organ systems failing, with only 1 of 10 patients surviving to ICU discharge with 3 or more organ system failures, compared with 31 of 47 patients surviving to ICU discharge with less than 3 organ system failures. 6 of the 7 patients diagnosed with veno-occlusive disease died.

2. How precise are the estimates of likelihood?

Again, no predictive factor was found to have 100% certainty in separating survivors from non-survivors. Of the 25 ICU deaths, 15 patients died after withdrawal of intensive care treatment and 5 had limits placed on escalation of therapy. Though the authors emphatically state that the majority of these twenty patients had no hope of recovery, the fact that 87% of deaths occurred after withdrawal or limitation of care could potentially obscure the natural course of illness and death in these patients.

(Confidence intervals were not reported in the article, but were calculated using the modified Wald method from http://graphpad.com/quickcalcs/ConfInterval1.cfm (1) and reported below. )

Outcome	Survival	Mean Survival (Calculated 95% C.I.)
ICU Discharge	32/57	56.1% (43.2%; 68.2%)
30-days post ICU DC	21/57	36.8% (25.5%; 49.9%)
6-months post ICU DC	11/40	27.5% (15.9%; 42.9%)

Paying attention to whether or not the survival data reported in the study was based on number of admissions or number of patients is also important in interpreting the study results. For instance, the overall 6-month mortality rate of 72.5% reported in the study was based on 11 survivors from forty patients, rather than 57 admissions. Using number of admissions, the overall mortality rate would have been 80.7% (giving a 6-month survival as shown in the table above of 19.3% (95% CI 10.9%; 31.5%)), reflecting multiple readmissions for patients that ultimately did not survive.

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

1. Were the study patients and their management similar to my own?

Yes. Great Ormond Street Hospital for Children is a tertiary care referral pediatric hospital with an active BMT unit capable of caring for severely ill children. Therefore admission to the ICU is usually based on the need or probability that a patient will require either inotropic or ventilatory support. The ICU uses a ventilation strategy based on permissive hypercapnea (PaCO2 < 60 mmHg & pH > 7.25), limitation of peak inspiratory pressure (< 35 cm H2O) and high mean airway pressure using a Servo 300 ventilator. Patients requiring > 60% FIO2 or a mean airway pressure > 20 cm H2O are usually treated with high-frequency oscillatory ventilation (HFOV) using a Sensormedics 3100A ventilator.

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

Yes and no. Despite the overall poor prognosis for critically ill children after BMT, no predictive factor was identified that was capable of identifying survivors from non-survivors with 100% certainty. Though patients requiring mechanical ventilation for longer than 48 hours due to a respiratory infection or those requiring HFOV carry a dismal prognosis, there are limited numbers within these subgroups to make absolute treatment decisions based upon them. (In fact, we can calculate the upper 95% confidence limit for the survival rate of 0% amongst those, for example, treated with HFOV with the simple rule-of-thumb of 3/N, with N in this case being 10 patients treated with HFOV. So although the measured survival rate was 0, the survival rate, with 95% confidence, could have been as high as 30%!(2))

These results are within the range of other reported BMT survival rates and therefore seem to collaborate that those previous results (although mostly in 1980'and 1990's) can still be applicable today despite our improved ICU treatment. It would have been nice to see much better numbers than the "olden" days of PICU.

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

Yes and no. As a framework for discussion, the results are useful in underscoring the fact that, despite the capabilities of modern intensive care, most pediatric bone marrow transplant recipients admitted to the ICU will die despite the best efforts of caregivers. These facts must be honestly shared with patients, families, and the ICU and primary care teams. However, the drive for basing these discussions on realistic expectations should not be construed as limiting hope, as there continue to be long-term survivors among these patients.

Intensivists keep looking for the "no-hope" constellation for these patients. That way they can appropriately counsel the families to limit support and avoid prolonged futile ICU admissions. However, there probably will never be a 100% sure predictive model for these patients.

References

1. Agresti A, Coull BA. Approximate is better than "Exact" for interval estimation of binomial proportions. The American Statistician. 1998; 52:119-126. For a description of the modified Wald method for determining 95% confidence intervals for a proportion. Read more ...
2. Hanley JA, Lippman-Hand A. If nothing goes wrong, is everything all right? JAMA 1983; 249: 1743-1745. [citation]

 

Comments

Back to the EB Journal Club Index