Transfusion in Critical Illness

Thanks to Dave and Heather Collins for a highly entertaining evening, and to J Beckemeyer, Catherine, Elise, Erin, J Remke and Dr. Febbo for their erudite presentations.

Increasingly, data support a more restrictive approach to blood transfusion.  Think of a blood transfusion as a liquid organ transplant, with added risks of TACO, TRALI and other transfusion related immunomodulation. 

BOTTOM LINE FOR JOURNAL CLUB:  7 is the new 10, and 1 is the new 2.  Transfuse at a hemoglobin of 7 g/dl unless unstable/actively hemorrhaging or having ACS, and give 1 unit of pRBCs at a time.  For ACS, current literature is inconsistent but reviewed article favored the more traditional approach of transfusing to 10 g/dl. 

1.     Villanueva C, et al. Transfusion Strategies for Acute Upper Gastrointestinal Bleeding. N Engl J Med 2013; 368:11-21.

In this single center RCT of 921 patients with acute upper GI bleeding, half were randomized to a restrictive strategy (transfuse when hemoblobin < 7 g/dl) and half to a liberal strategy (transfuse when hemoglobin < 9 g/dl).  Patients were included if >18 yo and evidence of acute upper GIB by hematemesis, melena, or both.   Patients with massive exsanguinating bleeding, ACS, symptomatic PVD, and TIA/CVA were excluded.  They also excluded patients with Rockall scores of zero and hemoglobin >12 (wimpy bleeds).  Randomization was stratified according to presence/absence of cirrhosis.  Transfusions were given 1 unit at a time, with reassessment.  Transfusion could also be given for symptomatic anemia, massive bleeding, or need for surgical intervention.  All patients underwent endoscopy within 6 hours.

Primary outcome of survival at 6 weeks was significantly higher in the restrictive (95%) versus liberal (91%) group.  Risk of further bleeding, adverse events, and rescue therapies were significantly lower in the restrictive group. Patients in the liberal group also had significant early rises in portal-pressure gradient (increases risk of rebleeding).  The only subgroup without a significant mortality difference were patients with advanced cirrhosis.

Why does restrictive strategy work in GI bleed?  Authors posit that transfusion may counteract protective splanchnic vasoconstrictive response caused by hypovolemia.  This increased pressure may impair clot formation.  Transfusion may also induce coagulation abnormalities, and for patients with cirrhosis, increased blood volume can increase portal pressure leading to rebleeding.

BOTTOM LINE:  great study looking at a population that makes us nervous, supporting a restrictive transfusion strategy to improve mortality, decrease rebleeding/adverse events, and conserve resources.  It’s a win/win, but remember that they excluded patients with exsanguinating bleeding as well as those with ACS.  Also, transfusion could be given for symptomatic anemia, massive bleeding, or if need for surgery-important caveats.  Finally, these patients all underwent early endoscopy, facilitating early intervention for ulcer, varices, etc.


2.     Holst LB, et al:  Lower versus higher hemoglobin threshold for transfusion in septic shock.  N Engl J Med  2014 9;371(15):1381-91. 

In this multicenter, randomized trial, 998 ICU patients with septic shock were randomized to a lower threshold group (transfuse 1 unit pRBCs for hemoglobin < 7 g/dl) or higher threshold group (transfuse 1 unit pRBCs for hemoglobin < 9 g/dl). 

Patients with stable cardiovascular disease were included, but patients with life threatening bleeding or ACS were excluded.  Patients developing life threatening bleeding, ischemia, or need for ECMO or surgery after enrollment could receive transfusion at discretion of treating physician.  Leukoreduced blood was used in an attempt to mitigate immunomodulatory effects of transfusion.

Primary outcome of death by 90 days occurred in 43% of lower threshold group compared with 45% of higher threshold group (RR 0.94; 95% CI 0.78-1.09).  Rates of ischemic events, severe adverse reactions, and life support requirements were also similar in the two groups.  The lower threshold group received half the total number of transfusions as the higher threshold group.  It’s an ICU, not an ED study, but given our boarding, probably still pretty useful information.

BOTTOM LINE: in ICU patients with septic shock, similar mortality and adverse event rates were observed with a lower transfusion threshold (transfuse for hemoglobin < 7 g/dl), with significant savings in the number of units of blood transfused.


3.     Carson JL, et al: Liberal versus restrictive transfusion thresholds for patients with symptomatic coronary artery disease. Am Heart J  2013;165(6):964.

Ok, so clear that in GI bleed and sepsis (and from the unfortunately named TRICC trial, for critically ill patients in general) a more restrictive blood transfusion strategy is safe and effective.  The holy grail remains...what about patients with ACS??

In this “pilot trial” of 110 patients with ACS or stable angina undergoing cardiac cath, patients were randomized to receive transfusion for hemoglobin < 10 g/dl (liberal strategy), or to transfusion for symptomatic anemia as well as permitted/not required to receive transfusion for hemoglobin < 8 (restrictive strategy).

Patients with active bleeding, hemodynamic instability, symptomatic anemia at time of randomization, or needing surgery were excluded.

Primary outcome was the composite of death, MI, or unscheduled revascularization 30 days after randomization, and occurred in 11% of liberal group and 25.5% of restrictive group (95% CI 0.7% – 29%).  Death at 30 days was 2% in liberal group vs. 13% in restrictive group (95% CI 1.5% – 21%).  These were not significant differences due to small group sizes/large CI, but trends favored liberal transfusion strategy.  All deaths were classified as cardiac.  Most other adverse cardiac outcomes were more frequent in restrictive group.

This study had problems.  First, it’s a small study, so can’t draw practice-changing conclusions.  Also, the restrictive group was on average 7 years older than the liberal group.  Patients with ACS are different from those with stable disease undergoing cath-really these are 2 separate populations.  The study was terminated prior to completing their planned enrollment of 200 patients...not clear why...they do some hand waving about having enough information to plan a larger trial.  Interestingly, the most frequent reason for protocol violation was insufficient time to administer transfusion prior to discharge in the liberal group (wait Mr. Smith...don’t call a taxi yet, let’s give you an [unnecessary?] transfusion).

BOTTOM LINE:  Not a great study, but trends support a more liberal strategy (transfuse for hemoglobin < 10) in patients with ACS. Existing literature for this population is inconsistent, equipoise exists, and a large, well done study still needs to be performed.

Sepsis Treatment Update

Many thanks to Mike and Lisa Anderson for hosting, and to Sola, Dr. Febbo, Jessica, Trale, Adam B. and Natalie H. for their erudite presentations.

Background:  The 2001 NEJM study by Rivers established a new standard for sepsis care, demonstrating a significant improvement in severe sepsis/septic shock mortality using a protocol driven care algorithm during the initial 6 hours of resuscitation.  Since then, there has been ongoing discussion regarding which components of the algorithm are the most impactful.

Bottom line:  Sepsis rates have decreased by nearly 50% since 2000, and this is likely a real finding, in part due to increased awareness and more aggressive treatment of sepsis, as well as to other improvements in critical care. The ProCESS trial reinforces the 4 most crucial elements of successful sepsis management:  early recognition, early and adequate IV fluids, early IV antibiotics, and clinical reassessment of circulation.

Article 1:  Kaukonen K, Bailey M et al.  Mortality Related to Severe Sepsis and Septic Shock Among Critically Ill Patients in Australia and New Zealand, 2000-2012. JAMA.2014;311:1308-1316.

Editorial response:  Iwashyna T, Angus DC.  Declining Case Fatality Rates for Severe Sepsis:  Good Data Brings Good News with Ambiguous Implications.  JAMA.2014;311:1295-1297.

This retrospective observational study and accompanying editorial describe the changes in mortality for severe sepsis with and without shock in 101,064 patients from 171 ICUs in New Zealand and Australia between 2000 and 2012.  The primary endpoint was hospital outcome (mortality and discharge home, to other hospital, or to rehab facility).  Absolute mortality in severe sepsis decreased over this time period from 35% to 18.4%, for an annual average decrease of 1.3%.  Interestingly, mortality also decreased at a similar rate in non-septic ICU patients over this time period.  The annual rate of discharge to home was significantly greater in severe sepsis patients compared to patients with non-sepsis diagnoses. 

As discussed in the editorial, reports of changes in disease incidence or mortality can be misleading and instead reflect increased disease awareness and more liberal testing.  This study demonstrated robust methodology by using consistent international consensus definitions of sepsis over the time frame of the study, prospectively gathered data, and by attempting to control for confounding variables, all lending strength and veracity to its conclusions.  This study points out the importance of taking into consideration overall improvements in care when evaluating new therapies, as a new drug’s “benefit” may be inflated due to an ongoing independent positive trend in outcomes.  Also, short-term mortality by itself is an inadequate outcome metric, as a reduction in short-term mortality may come at the expense of trade-offs in morbidity and mortality.  See Katie Burns for diagrams and an interpretive explanation.


Article 2:  The ProCESS Investigators.  A Randomized Trial of Protocol-Based Care for Early Septic Shock.   NEJM March 18,2014. 

Editorial response:  Lily CM: The ProCESS Trial-A New Era of Sepsis Management. NEJM. March 18, 2014.

ACMC was an enrollment center for the ProCESS Trial, a non-blinded RCT of 1341 patients with septic shock, comparing 3 arms of care:  EGDT (Early Goal Directed Therapy, Rivers algorithm with mandated central line and central hemodynamic monitoring, specific physiologic targets); less aggressive protocol based care (no mandated central line, emphasis on clinical assessment of perfusion); or treating doctor’s “usual care.”  The primary end point was 60 day in-hospital mortality.  Secondary outcomes included longer-term mortality and organ failure/need for organ support.

At 60 days, mortality was 21% in the EGDT group, 18.2% in the less aggressive protocol group, and 18.9% in usual care group.  Protocols were not superior to usual care, and the EGDT protocol was not superior to the less aggressive protocol.  There were also no significant differences in any secondary outcomes, except for a higher need for new dialysis in the less aggressive protocol group (6%) than in the EGDT (3.1%) and usual care (2.8%) groups.

In the initial 6 hours of care, IVF and pressor rates did vary between groups (most IVF in the less aggressive protocol group, more pressors in both protocol groups than in usual care group).  More patients in EGDT group than in other 2 groups received inotropes and blood transfusions.  Antibiotics and steroid use were similar across all groups.

Discussion at JC:  The types of fluid/pressor were not specified, and neurologic outcome was not measured (although percentages of patients discharged to home were similar in the 3 groups).  Patients in both of the protocol groups basically had their own resuscitation doctor for 6 hours as emergency physicians were on call to come to the hospital and provide dedicated care for these 2 groups of patients during the initial 6 hours of the study.  This may reflect even better on the usual care group’s outcomes.  For Febbo, interesting that the group receiving the most crystalloid also had the highest renal failure rate....hyperchloremic metabolic acidosis from high volume Normal Salineàrenal failure??

Although institutions enrolling patients in ProCESS were not supposed to be actively using a sepsis protocol prior to becoming involved in the trial, 70% of participating institutions did have an existing sepsis protocol.  All groups in ProCESS received on average more than 2 liters of IVF prior to randomization, and 76% of patients received antibiotics prior to randomization, reflecting knowledge of EGDT in physicians treating “usual care” patients.

The 18% mortality rate in the “usual care” group of ProCESS is a dramatic improvement in baseline mortality compared with the 46.5% control group mortality in Rivers’ trial.   General improvements in critical care have also likely been impactful, including more liberal transfusion thresholds, lung protective ventilation strategies, and moderate glucose control.

So, do we need sepsis protocols at all?  Critical care resuscitation is complex.  Especially for residents and physicians with infrequent experience treating sepsis, protocols provide useful reminders of treatment goals and therapies.   Moving forward, the traditional EGDT protocol will likely evolve, with less pressure to insert central lines and measure CVP/ScvO2 unless the patient requires vasopressors.  However, reassessing perfusion using serial lactates, Shock Index, and clinical parameters including urine output and mentation remains critical.