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:
Impact of allogenic packed red blood cell transfusion on nosocomial infection rates in the critically ill patient.
Taylor RW, Manganaro L, O'Brien J, Trottier SJ, Parkar N, Veremakis C.
Crit Care Med. 2002;30(10):2249-54.
[abstract]
Reviewed by Renán Orellana, MD
, Baylor College of Medicine, Pediatric Critical Care Section,
Texas Children's Hospital, Houston, TX
Review posted February 8, 2003
I. What is being studied?:
- The study objective:
To determine the association between allogenic packed red blood cells (PRBC) transfusions and the incidence of nosocomial infection rates in a tertiary hospital.
- The study design:
This was a retrospective cohort analysis, but there was no temporal dimension to the study; the data were analyzed more like a cross-sectional survey than a cohort study.
- The patients investigated:
1717 adult patients admitted to a single surgical-medical trauma ICU recorded in the Project IMPACT database from 10/1/98 to 8/31/00. The nosocomial infection (NI) rates were determined and subjects were classified according to PRBC transfusion. The definition of NI followed the specific operational criteria for NI observed by Project IMPACT. These are not defined in the article.
- The patients excluded:
No exclusion criteria reported.
- The treatment, intervention, or exposure in question :
Allogenic PRBC transfusions.
- 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 group (patients that did not receive PRBC) and the exposed group (patients that received at least one unit of PRBC) were not clearly identified; the only patient characteristics provided were probability of survival (POS) from the MPM-0 tool and age. These are outcomes, not pre-existing patient demographics that would allow a valid comparison between both groups. We do not know the similarity between controls and transfused patients in all other important confounders or demographic characteristics before the transfusion exposure, such as diagnoses, invasive devices, immune status and their temporal association with the appearance of the NI. LOS, also an outcome, was significantly longer in the transfused group; however, it is not clear that this occurred because of blood transfusions, or because they had a more severe illness. It is possible, for example, that the transfused group had twice as many penetrating trauma patients as the non-transfused group; the authors claim that with 49 different diagnoses, no inferences could be made.
Clearly patients could have been further characterized. In fact, there is no typical "Table 1" - the table that demonstrates the numerous demographic factors in the two groups (the Table 1 in this paper is a review of blood transfusion risks culled from the literature). This lack of accounting for such important variables seriously jeopardizes this study's validity.
- 2. Were the exposures and outcomes measured in the same way
in the groups being compared?
Yes. Recollection, recall, and surveillance bias was avoided by using data already recorded that utilized specific operational definitions. We presume the data collectors were not aware of this study's hypothesis and unlikely to record NI differently in those transfused and those not transfused.
- 3. Was follow-up sufficiently long and complete?
Yes. Recorded data followed patients until discharge or death.
- Secondary questions:
- 4. Is the temporal relationship correct?
They did not specify the temporal relationship between blood transfusions and NI. It is unclear how long after - or before - the transfusion the reported NI occurred.
- 5. Is there a dose-response gradient?
Yes, the authors report a dose-response gradient in the rate of NI depending on the units of PRBC's transfused. The more units transfused, the greater chance of infection. However, the data to support this finding are not shown in the paper.
- 1. How strong is the association between exposure and outcome?
The authors chose an odds ratio to report the strength of association between PRBC transfusion and NI rates. According to the data presented, a nosocomial infection is 6 times more likely to develop in a patient who receives PRBC transfusion than a patient who does not. The odds ratio was calculated as follows:
| NI | No NI | Totals |
PRBC TRANSFUSION | 64 | 352 | 416 |
No Transfusion | 38 | 1263 | 1301 |
OR = [(64)(1263)]/[(38)(352)] = 6
Since this is a cohort study, not a case-control, the estimation of Relative Risk (RR) could have been calculated [RR = (64/416) / (38/1301) = 5.26]. The absolute risk increase can be estimated (ARI = CER (control event rate) - EER (exposed event rate) = 15.38 - 2.92 = 12.46 ∼ 12%). However, given the lack of any temporal evidence that transfusions preceded infections, calculation of risk in this fashion has tenuous validity.
- 2. How precise is the estimate of the risk?
The confidence intervals required to assess the accuracy of the odds ratio are 4 and 9.2. Thus, we can say with 95% certainty that the OR of the association between a NI and a PRBC transfusion is between 4 and 9.2.
- 1. Are the results applicable to my practice?
Yes. The PRBC transfusion is part of a daily practice in critical care. However, it would be necessary to adjust to the confounders already mentioned and establish a temporal relationship between transfusion and the outcome to prove that the magnitude of the reported effect is accurate. From this study we know only that there is an association between transfusion and nosocomial infection; there is no evidence for cause and effect from this study.
- 2. What is the magnitude of the risk?
The number needed to harm (NNH), calculated from the odds ratio, suggests that for every 8 patients who receive a blood transfusion, one will develop a nosocomial infection. For the odds ratio, the NNH was calculated as follows (PEER: patient expected event rate not exposed):
NNH | = | PEER (OR - 1) - 1
[PEER (OR - 1) - 1] X (1 - PEER) |
| | | | |
NNH | = | 0.0292 (5) - 1
[0.0292 (5) - 1] X (1 - 0.0292) |
| | | | |
NNH | = | 8 | | |
If we considered this a true cohort study, the NNH can be calculated as the reciprocal of the ARI: 1/12 = 8. The 95% CI for the ARI is from 9 to 16%, and from their reciprocals, we can say with 95% certainty that for every 6 to 11 patients who receive a blood transfusion, one will develop a NI.
- 3. Should I attempt to stop the exposure?
The authors report a significant increase in the magnitude of the risk of developing NI by current PRBC transfusion practices. Although the strength of the study is challenged by unreported confounders, this robust association is supported by a landmark randomized controlled trial (1). Some patients may require red blood cell transfusion because of their critical condition; as an alternative, lowering the transfusion practice threshold to a level that is tolerated may benefit the patient by reducing the risk of NI and organ failure (1,2). However, before attempting to reduce current PRBC practice, it is important to determine the risk/benefit of PRBC transfusions. Therefore, the risk of not transfusing PRBC need to be established.
References
- Hebert PC, Wells G, Blajchman MA, Marshall J, Martin C, Pagliarello G, Tweeddale M, Schweitzer I, Yetisir E. A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care. Transfusion Requirements in Critical Care Investigators, Canadian Critical Care Trials Group. N Engl J Med 1999 Feb 11; 340 (6):409-17. [abstract; full-text for subscribers]; [PedsCCM EB Journal Club review]
- Vincent JL, Baron JF, Reinhart K, Gattinoni L, Thijs L, Webb A, Meier-Hellmann A, Nollet G, Peres-Bota D; ABC Anemia and blood transfusion in critically ill patients. JAMA. 2002 288(12): 1499-507. [abstract]
Comments
Submit comments regarding this review by e-mail
or
with the EB
Journal Club Comment Form
Back
to the EB Journal Club Index
|