What's Your Trigger?

The Gist: "Permissive _____" is becoming increasing popular in medicine-blood glucose, blood pressure, and oxygen saturation. Hemoglobin is similar, but common refrains may be heard "well, he looks puny, let's give him a couple of units." Currently, the best evidence suggests that transfusion of packed red blood cells (PRBCs) should be considered in most critically ill medical patients, in the absence of massive hemorrhage, at a hemoglobin (hb) <7 g/dL. The data show that liberal transfusions don't benefit the patient and may harm them. However, there are limitations to the data and it and should be interpreted within the context of the individual patient. Treat the patient, not the lab value! Prevent iatrogenic anemia, resulting in more transfusions.  Excellent review article from Annals of Intensive Care

The case(s): In the ED, transfusion of PRBCs is often clear-cut. Massive hemorrhage with high TASH score? Activate the protocol. Patient with melena, as white as the sheets, with a hemoglobin of 4.0 g/dL? Hang the blood. What about the 60 year old patient with suspected sepsis, resting comfortably without complaints, who has a hemoglobin of 7.8 g/dL (baseline ~8-8.5 g/dL)? What if we know he is undergoing fluid resuscitation due to the sepsis, borderline hypotension, and tachycardia, which will likely cause a dilutional drop in hemoglobin?
I recently encountered similar cases and found the art of medicine plays a large role when patients. Tintinalli recommends transfusion at 7.0-9.0 g/dL, but our patient, like many, falls within that grey area so any course of action is justified (Ch 146). Can the literature help sort this out?

What did the FOAM say?

• The NNT review (2010)- 100% saw no benefit, 1 in 18 harmed by pulmonary edema by liberal transfusion.
• Yasser Sakr's talk offers some skepticism to this practice (note: second author on the SOAP study paper).  

How About the Society Guidelines?
Society of Critical Care Medicine.  Suviving Sepsis 2012 Guidelines

• RBCs when the hemoglobin concentration decreases to < 7.0 g/dL (target a hb 7.0-9.0 g/dL)

AABB (American Association of Blood Banks) Red Blood Cell Transfusion: A Clinical Practice Guideline From the AABB (full text)

• Consider transfusion at hb of 7 g/dL or less. 
• Postoperative surgical patients, consider RBCs at hb of 8 g/dL or less or for symptoms (chest pain, orthostatic hypotension or tachycardia unresponsive to fluid resuscitation, or CHF)

American College of Critical Care Medicine Clinical practice guideline - Pulm CCM review

• Transfuse RBCs as single units in the absence of hemorrhage (Level 2 evidence)
• Cognitive changes seem to occur at <5 g/dL so some asymptomatic, hemodynamically stable patients may not need transfusion with hb 5-7 g/dL. 

What do we mean by symptoms?  

• Cognitive changes, syncope, dyspnea, chest pain, etc

Transfusion Problems.. FOAM resources, FOAMcast, LITFL review and Gould et al)

• Febrile nonhemolytic reactions (most common)
• Hemolytic transfusion reaction (type 2 hypersensitivity)
• Transmission of pathogens (Viral, Bacterial)
• TRALI (transfusion-associated lung injury), Pulmonary Edema, ARDS
• Transfusion Associated Circulatory Overload (new respiratory distress and hydrostatic pulmonary edema within 6 hours after RBCs) - associated with renal failure and number of units (ref)
• TRIM (Transfusion Related Immunomodulation)
• Biochemical - may lead to vasoconstriction, GFR changes, hyperkalemia (older cells)
• Hypothermia, coagulopathy (dilutional), thrombocytopenia with massive transfusion.
• Human error
• Expensive -Cost to transfuse 1 unit PRBCs to patient $1600-2400 (ref)
 

Prevention:

•  Avoid Iatrogenic Anemia.  Spahn et al address this artfully.

Literature base for society recommendations: 

• Based on a single RCT (TRICC), and a few observational studies
• These all look at a hemoglobin level as a transfusion trigger, not symptoms.
• Observational studies have large potential for bias due to variation in a physician's decision to transfuse as well as individual patient factors (despite statistical models to account for this).

Cochrane review 2012

• 19 trials (Diverse: 8 surgical, 5 in setting of acute hemorrhage, 1 oncologic, 3 critical care, 1 pediatric), n=6264 patients
• 30 day mortality - no significant difference (RR 0.85, 95% CI 0.70-1.03)
• In hospital mortality - lower in restrictive group (RR 0.77, 95% CI 0.62-0.95) 
• Risk of receiving RBCs - average absolute risk reduction of 34% (95% CI 24%-45%). The volume of RBCs transfused was reduced on average by 1.19 units (95% CI 0.53-1.85 units).
• Restrictive transfusion strategies did not appear to impact on the rate of cardiac events, myocardial infarction, stroke, and thromboembolism (i.e. appeared as safe from a hemodynamic standpoint).
• Infection - no significant difference (6 trials)
• Limitations - Lots of heterogeneity, TRICC study contributed the majority to the review.

Trial of Transfusion Requirements in Critical Care (TRICC trial) Hebert et al

• Multicenter RCT, n=838, ICU setting.  Restrictive: transfuse <7.0g/dL; Liberal transfuse <10.0 g/dL
• 30 day Mortality - no statistical difference between groups: 18.7% vs 23.3% in the restrictive vs. liberal-strategy group (95% CI –0.84-10.2%)
• Subgroup analysis showed that younger patients (<55 years old) and less sick patients (APACHE II <20) who received transfusion had increased mortality.
• Mortality rates during hospitalization- lower in the restrictive group 22.2% vs. 28.% (P=0.05)
• ICU mortality- lower in restrictive group, but not significant 13.9% vs. 16.2% (P=0.29) 
• 60-day mortality- lower in restrictive group, but not significant 22.7% vs. 26.5% (P=0.23)
• Limitations:  
• Study stopped early due to poor enrollment. Physicians hesitant to enroll patients, possibly due to fear of restrictive transfusion.
• Excluded: chronic anemia, surgical patients, active bleeding.
• Subgroup analyses: underpowered.

Anemia and blood transfusion in critically ill patients (ABC study) Vincent et al (2002)

• Observational study, European ICUs n=3534 patients
• ICU  mortality =18.5% vs 10.1%, (transfused vs not transfused); P<.001
• Overall mortality rates =29.0% vs 14.9%, P<.001 (transfused vs not transfused)
• For similar degrees of organ dysfunction, patients who had a transfusion had a higher 28-day mortality rate 22.7% vs 17.1% (p =.02) in a matched patients in the propensity analysis
• Higher mortality in ICU patients receiving PRBC transfusion (OR of death of 1.37).
• Limitation: observational, although analysis accounted for degree of organ dysfunction, provider discretion played a role in decision to transfuse; thus, the sicker patients likely received more transfusions.

The CRIT study Corwin et al (2004)

• Prospective, multi-center, observational study in US ICUs with n = 4,892
• Upon mulivariate analysis, the number of PRBCs a patient received was independently associated with longer ICU and hospital LOS and an increase in mortality 1.65; 95%CI 1.35–2.03, p <.001). Note: one model showed increased mortality in patients with hb nadir <9.0.
• Patients who received transfusions also had more total complications.
• Most common reason for transfusion - low hemoglobin (mean transfusion hb = 8.6 g/dL) not symptoms
• Excluded: cardiac/burn/neuro/pediatric ICUs, renal failure. Limited by statistical analysis instead of direct comparison.

SOAP study Vincent et al (2008)

• Prospective, multi-center, observational. n=3,147 patients
• Transfused patients had higher ICU LOS (5.9 vs. 2.5 days; P <0.001)
• Transfused patients had higher ICU mortality rate (23.0 vs. 16.3%; P<0.001)
• Lower 30-day hazard of death in the transfused group when adjusted (HR 0.73; 95% CI 0.59–0.90; P <0.004)
• Why the difference from the very similar ABC study?  
• Used more leukodepleted blood than the ABC study. This is now standard practice.
• More knowledge about potential harms of transfusions by this time (ABC, TRICC studies already published) so more likely that sicker patients received the transfusions.

The evidence is basically non-inferior from a morbidity and mortality standpoint for the use of 7.0 g/dL and presence of symptoms as a transfusion trigger in medical patients without acute hemorrhage.

Update 2014:  Holst et al performed a randomized, multicenter, parallel-group study (TRISS) in which patients with sepsis were randomized to transfusion triggers <9 g/dL or <7g/dL.  In another demonstration that transfusion at a hemoglobin <7 g/dL in patients without active ischemia does not incur excess harm, the study found that 90 day mortality was not significantly different between the groups (43% <7 g/dL and 45% <9 g/dL).