Correspondence

Blood Transfusions in Critical Care

N Engl J Med 1999; 341:123-124July 8, 1999

Article

To the Editor:

In their multicenter trial of transfusion requirements in critical care, Hébert et al. (Feb. 11 issue)1 demonstrate that, in the case of red-cell transfusion, the adage “more is better” is not always true. However, their recommendation of a threshold for red-cell transfusion of a hemoglobin concentration of 7.0 g per deciliter may not be appropriate at high altitudes.

Although a hemoglobin concentration of 7.0 to 9.0 g per deciliter may suffice at an atmospheric pressure of 760 mm Hg at sea level, it yields a dangerously low partial pressure of arterial oxygen at a higher altitude (such as we have here in Santa Fe), and even in relatively healthy patients, these values may not be safe. Those who live at high altitudes routinely maintain base-line hemoglobin concentrations substantially higher than those of their relatives living at sea level, and inadequate oxygen content and delivery could unmask underlying cardiac or other disease. Readers should be aware that the target hemoglobin concentrations recommended by Hébert et al. may need to be tailored according to the altitude of their institutions and that the use of supplemental oxygen is warranted in patients being treated at high altitudes.

Michael A. Pilla, M.D.
St. Vincent Hospital, Santa Fe, NM 87505

1 References

1. 1

   Hebert PC, Wells G, Blajchman MA, et al. A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care. N Engl J Med 1999;340:409-417
   Full Text | Web of Science | Medline

To the Editor:

In their study of transfusion requirements in critically ill patients, Hébert et al. performed subgroup analyses that support their conclusion, with the possible exception of the analysis of patients with myocardial infarction and unstable angina. However, no subgroup analysis was performed for patients with primary or secondary brain injury (those with traumatic brain injury or subarachnoid hemorrhage or those who have undergone neurosurgery). These patients may be another exception to the rule of restricting transfusion, since the occurrence of a low ratio of cerebral oxygen delivery to cerebral metabolic rate for oxygen adversely affects the outcome after brain injury.1 A correct hemoglobin concentration is crucial in the struggle to optimize cerebral oxygen delivery in this setting. Although no clear-cut transfusion trigger has been proposed for patients with brain injury, neurointensivists usually adopt the liberal strategy of maintaining hemoglobin concentrations at 10 g per deciliter.

Marco Gemma, M.D.
Luigi Beretta, M.D.
Scientific Institute S. Raffaele Hospital, 20132 Milan, Italy

1 References

1. 1

   Marion DW, Darby J, Yonas H. Acute regional cerebral blood flow changes caused by severe head injuries. J Neurosurg 1991;74:407-414
   CrossRef | Web of Science | Medline

To the Editor:

Hébert et al. report that “an average of 2.6±4.1 red-cell units per patient was administered in the restrictive-strategy group, as compared with an average of 5.6±5.3 units per patient in the liberal-strategy group (P<0.01).” Did the authors use the correct statistical method to evaluate such a skewed distribution? Did they analyze the results on the basis of the number of units of blood used? Is there a cutoff point related to imminent multiple-organ dysfunction?

William J. Ledger, M.D.
New York Presbyterian Hospital, New York, NY 10021

To the Editor:

We congratulate Hébert et al. on their timely and important study of the transfusion needs of critically ill patients. We agree with the authors' supposition that arbitrarily designated transfusion triggers (e.g., 10 g per deciliter) are inherently wrong, but we believe their study does not support this assumption. The authors studied all patients admitted to the intensive care units of the participating institutions without stratification according to disease or severity of disease. Acute Physiology and Chronic Health Evaluation (APACHE II) scores group together patients with very different outcomes. For example, a patient with multiple trauma may have a low APACHE II score, as compared with a patient with coma (but no other underlying disorders), who may have a high score; both patients would be included in this study.1 Did the patients with respiratory or cardiac disease have acute exacerbations or chronic disease? Clinical experience suggests that patients with these disease profiles tolerate oxygen debt differently.2

Our greatest concern is the external validity of the study. Why did only 13.0 percent of the group of patients initially screened ultimately undergo randomization? Why did so many physicians withhold consent for such a large number of eligible patients? How did the group for which the physician's consent was withheld (598 patients) differ from the 838 randomized patients in terms of base-line characteristics, transfusion levels, and outcomes?

The authors describe a rate of noncompliance among physicians that translates to 24 patients (18 of 420 patients in the liberal-strategy group and 6 of 418 in the restrictive-strategy group); however, in the only statistically significant outcome with respect to mortality, the difference in in-hospital mortality was 25 patients. The group with noncompliant physicians may obscure the difference between the restrictive-strategy group and the control group, potentially negating the findings of the study. The poor adherence to the transfusion protocol after patients left the intensive care units, the low odds ratio, and the lack of indexes of interrater reliability all weaken the results.

Paul Barach, M.D., M.P.H.
Stephen D. Small, M.D.
Massachusetts General Hospital, Boston, MA 02114

2 References

1. 1

   Rogers J, Fuller H. Use of daily Acute Physiology and Chronic Health Evaluation (APACHE) II scores to predict individual patient survival rate. Crit Care Med 1994;22:1402-1405
   CrossRef | Web of Science | Medline

2. 2

   Asmundsson T, Kilburn KH. Survival of acute respiratory failure: a study of 239 episodes. Ann Intern Med 1969;70:471-485
   Web of Science | Medline

Author/Editor Response

Dr. Hébert replies:

To the Editor: My colleagues and I believe that Dr. Pilla is confusing a lowered partial pressure of oxygen with a decrease in oxygen transport. Given that 99 percent of oxygen is transported bound to hemoglobin, a partial pressure of oxygen greater than 60 mm Hg (90 percent saturation) will usually ensure adequate delivery of oxygen.

Drs. Gemma and Beretta suggest that a liberal transfusion strategy is necessary to maintain adequate delivery of oxygen to damaged cerebral tissue. However, transfusions may decrease oxygen delivery to tissue by adversely affecting microvascular flow or the release of oxygen from hemoglobin (a shift to the right of the oxyhemoglobin-dissociation curve), an effect that results from the use of red cells after prolonged storage. In the small subgroup of patients with either closed head injury or subarachnoid hemorrhage (71 patients), there were no obvious benefits of transfusion (P>0.5). In short, we do not believe that patients with either type of injury should be exempt from our recommendations.

Dr. Ledger correctly notes that there were large variations and skewed distributions in the number of units transfused; even so, distributions were statistically different in both parametric and nonparametric analyses (P<0.01). Attempting to decipher the effect of transfusion on the incidence of organ failure and death would be analytically complex and subject to confounding factors.

Drs. Barach and Small state that the study population was heterogeneous, with a wide variety in terms of the severity of illness and in diagnostic groups. We suggest that the use of a heterogeneous patient population improves the generalizability and strength of the findings in a study in which strong trends or differences have been observed. In terms of tolerance of oxygen debt, there were no subgroups in the study that appeared to do better when treated according to a liberal transfusion strategy.

In addition, a large, randomized clinical trial is expected to improve substantially on the anecdotal experience of any one physician. An efficacy analysis, in which we excluded patients who were crossed over to the other treatment strategy or whose treatment was found not to be in compliance with the protocol, did not alter the results of the study overall or in clinically important subgroups. Adherence to the transfusion protocols once the patients had been discharged from the intensive care unit was not feasible. We also reasoned that the benefits and risks of red-cell transfusion would be most important during the early stages of critical illness rather than during recovery. Finally, ascertainment of objective outcomes, such as survival status and rates of organ failure, does not require assessment of interrater reliability, and the differences in outcomes, especially within subgroups, were sufficiently large for us to conclude that both treatments were not equivalent.

Paul C. Hébert, M.D.
Ottawa General Hospital, Ottawa, ON K1H 8L6, Canada

for the Transfusion Requirements in Critical Care Trial

Citing Articles (1)

Citing Articles

1. 1

   Karen K. Giuliano, Maria E. Grant. (2002) Blood Analysis at the Point of Care: Issues in Application for Use in Critically Ill Patients. AACN Clinical Issues: Advanced Practice in Acute and Critical Care 13:2, 204-220
   CrossRef