HARM

Criteria abstracted from The Users' Guides to the Medical Literature series in JAMA

Chest 2007 132:1116-1123..[abstract]

Review posted June 29, 2010

1. What is being studied?

   The study objective:

   The study objective was to evaluate the association between the transfusion of packed RBCs and mortality in patients with ALI/ARDS.

   The study design:

   The study design was a single center cohort study with the retrospective collection of transfusion data.

   The patients investigated:

   The patients investigated in the study were 248 patients with ALI/ARDS between 1999-2002. All patients >13 years old admitted to the medical or surgical intensive care unit at the University of Pennsylvania were screened for ALI/ARDS.   Those meeting the criteria for ALI/ARDS as defined by the American European Consensus Conference were enrolled within 48 hours of the onset of ALI/ARDS.   262 patients met eligibility criteria and were enrolled.   248 of these patients had full information on in hospital mortality and transfusion data.

   The criteria used to exclude patients may have been determined by the initial NIH study from which the patients were identified.  The exclusion criteria were if patients had a current or prior history of congestive heart failure, respiratory disease, or a condition that was similar to ALI/ARDS (vasculitis with diffuse alveolar hemorrhage, burns >30% of total body area), or had received a bone marrow or lung transplant. 

2. 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?

      The comparison groups were clearly identified as those who received blood transfusions and those who did not. Due to the majority of patients in the study having received a transfusion, analysis was also done in order to examine the effect of each blood transfusion.  Data regarding the baseline characteristics of patients who received transfusions versus those who did not or even information regarding groups who received a small amount of transfusions versus those who received a large amount of transfusions was not done.

      There were clearly multiple confounders in this study. The authors identified and evaluated the possible confounders of age, male gender, trauma, severity of illness (by apache III score), total length of ICU stay, total length of stay, long term alcohol use, and diabetes. The patients who died were statistically significantly more likely to be older, not have trauma as the source of their lung disease, have a higher APACHE III score, stay for a shorter period of time in both the ICU and the overall stay, have a history of long term alcohol use, and have diabetes. Diabetes, trauma as the etiology of ALI/ARDS and a history of long term alcohol use did change the point estimate of the OR at the threshold of > or equal to 15% and were therefore included in the multivariate models. 

      The authors decided to use only the APACHE III score as the marker of severity of other comorbidities due to the similarity between this score and other individual markers of disease such as metastatic cancer, cirrhosis, etc.  Due to concerns about colinearity the APACHE III score was used as the variable representing these other factors in the multivariable model.  APACHE III scoring does include adjustment for the co-morbid conditions of AIDS, hepatic failure, lymphoma, metastatic cancer, leukemia/multiple myeloma, immunosuppression, and cirrhosis. 

      During the study period, the blood bank transitioned to leukoreduced PRBCs and the ARDSnet study demonstrating the advantage of using a low tidal volume strategy was published.  The authors attempted to account for theses confounders of decreased mortality with multiple methods evaluating for changes over time with all evaluations not changing the OR for mortality.  It would have been helpful however to have evaluated as a possible confounder the tidal volumes that the patients in both groups received however it is possible this information was not available.  

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

      The primary exposure variable, packed RBCs, was evaluated as both a dichotomous variable (any transfusion) and linear variable (total number of packed RBCs transfused adjusted for length of stay to avoid immortal time bias).   Due to the blood bank at the institution changing to leukoreduced blood products during the study period, leukoreduced and non-leukoreduced products were evaluated separately as risk factors.  In addition, transfusion prior to the diagnosis of ALI/ARDS versus transfusion after the diagnosis of ALI/ARDS were evaluated as risk factors.   The primary outcome was in hospital mortality. 

      The exposure and outcome variables were measured in the same way in the groups being compared both had data obtained retrospectively from the blood bank transfusion records without knowledge of the ALI/ARDS outcome.  The data on outcome was collected as part of the National Institutes of Health, National Heart, Lung, and Blood Institute Specialized Centers of Research study in ALI/ARDS which lends credibility to the outcome data. 

   3. Was follow-up sufficiently long and complete?

      Follow up to hospital discharge seems appropriate.  The main mechanisms theorized to cause increased mortality after blood transfusion in acute lung injury is related to both immunomodulation and by amplifying lung injury through increased microvascular sludging and permeability.  It is possible that an effect could be seen after hospital discharge as red cells generally last around 120 days, transfused cells likely even less.  If the patient is immunosuppressed from the transfusion and more susceptible to infection this could be a possible outcome related to transfusion that was missed.  However, it is logically rather unlikely that a significant population of patients would improve to discharge then decompensate again. 

   Secondary questions:

   4. Is the temporal relationship correct?

      Yes, the outcome of interest, death, did not occur prior to the exposure of RBC transfusion.

   5. Is there a dose-response gradient?

      Yes, in a multivariable model OR for death per unit transfused was 1.06 (95% CI 1.04 to 1.09  p < 0.001).  This dose response gradient strengthens the association between RBC transfusion and death.  Evaluating whether this relationship was truly linear per unit transfused or if there was a change in risk at a certain unit transfused may have been helpful as well.

3. What were the results?
   
   
   1. How strong is the association between exposure and outcome?

      Patients who died were more likely to have received a PRBC (92% versus 78% p = 0.004) or a platelet (56% versus 41% p = 0.017) transfusion.  When transfusion of PRBCs was evaluated as a dichotomous variable the transfusion of any RBCs (greater than or equal to one unit) was associated with an unadjusted OR for mortality of 2.9 (95% CI 1.32-6.35, p = 0.008) compared to those not receiving RBCs.    When age, gender, APACHE III score were built into a logistic regression model this remained significant with an OR of 3.12 (95% CI 1.28-7.58, p = 0.012). 

      When length of stay was adjusted for in a multivariable model OR per unit transfused 1.06 (95% CI 1.04 to 1.09, p < 0.001). 

      The authors evaluated for the effect of transfusion both prior to and after the development of ALI/ARDS on mortality.  Transfusion of blood after the onset of ALI was associated with an OR for mortality of 1.13 (95% CI 1.07 to 1.20,  p < 0.001) per unit transfused in the multivariate analysis while administration of blood prior to ALI was not associated with mortality.  The lack of association with mortality for transfusion prior to the onset of ALI/ARDS misses those patients who may have died from transfusion prior to the onset of ALI/ARDS.  

      The authors also evaluated the effect of the change from non-leukoreduced PRBCs to leukoreduced PRBCs.  The OR for non-leukoreduced packed RBCs per unit transfused in the multivariate model (1.14 95% CI 1.07-1.21, p < 0.001) was higher than that for leukoreduced RBCs (OR 1.06 95% CI 1.03-1.09, p < 0.001).   It is not surprising that non-leukoreduced products with the inflammatory effect of white blood cells were associated with a higher mortality. 

      Although the OR are low, they are statistically significantly above 1 even in multivariate models and the OR are calculated per unit of blood.   Since many patients receive more than one unit of blood, this makes even a low OR notable.

   2. How precise was the estimate of the treatment effect?

      The OR results for the multivariate analysis of in hospital mortality per unit of PRBCs transfused, per unit of PRBCs transfused after ALI onset, and per unit of PRBC’s transfused that were leukoreduced or non-leukoreduced were all presented with 95% CI that were statistically significant, rather narrow, and did not cross 1.  

4. Will the results help me in caring for my patients?
   
   
   1. Are the results applicable to my practice?

      The inclusion criteria ( > 13 y/o) allowed for some pediatric patients to be included in the study however the baseline ages for each group are well above the pediatric range and most patients were not in the pediatric range.  It is possible that pediatric patients have a different immune response to the transfusion of PRBCs.   The association of mortality with PRBC transfusion in pediatric ALI has not been studied to my knowledge.  Previous studies have demonstrated that in pediatric critical illness, a low transfusion threshold for hemodynamically stable patients does not cause increase adverse outcomes (1).

   2. What is the magnitude of the risk?

      In this study with multiple confounders the magnitude of the risk was the adjusted OR for mortality of 1.06 per unit of packed RBCS transfused.   While this is a low OR, with each transfusion received it increases.

   3. Should I attempt to stop the exposure?

      There clearly seems to be an association between the transfusion or RBCs and mortality. The authors made an effort to control for possible confounding variables, however it is still possible that there are additional confounding variables that were not evaluated.   For example, platelet transfusion was evaluated as a confounder, however plasma was not evaluated.   Another possible confounder is the age of the blood transfused; this may make a difference in terms of deformability of red cells or cytokine levels in the blood transfused.  As seen in the study, non-leukoreduced PRBC’s seemed to have a higher risk associated with them; older blood products may also have a greater risk.   It is known that patients with direct lung injury versus indirect lung injury have a better outcome (2).   Although the authors evaluated the etiology of trauma as a possible confounder, the presence of another indirect etiology such as sepsis that increases mortality was not evaluated.

      Even knowing that possible confounders were not evaluated and this study was not done in pediatric patients, the evidence in this paper adds to the wealth of information from other studies regarding transfusion strategies in critically ill patients.  Based on these studies a conservative approach to transfusion seems prudent perhaps even more so in patients with lung injury.  

References:

1. Lacroix J, Hebert PC, Hutchison JS, Hume HA, Tucci M, Ducruet T, et al. Transfusion strategies for patients in pediatric intensive care units. N.Engl.J.Med. 2007 Apr 19;356(16):1609-1619.
2. Flori HR, Glidden DV, Rutherford GW, Matthay MA. Pediatric acute lung injury: prospective evaluation of risk factors associated with mortality. Am.J.Respir.Crit.Care Med. 2005 May 1;171(9):995-1001.

Last Updated: June 29, 2010

Fill out our Feedback Form | Editorial Notes | About Us