Perhaps the most contentious debate in critical care is whether and when to transfuse blood to patients, especially those with acute lung injury and/or septic shock. FACTT showed less fluids (which could include blood) are better for ALI/ARDS, but transfusion wasn't controlled and its contribution to the outcomes is unknown. Practicing physicians vary widely in their acceptance of RBC transfusion's place in early goal-directed therapy for severe sepsis -- with 30% of U.S. physicians saying they don't comply.
So ARDSNet crunched some numbers. Parsons et al parsed the original FACTT data and found all the patients who had septic shock in the first 24 hours (n=285). They further identified those who "should have" gotten blood early after randomization as part of early goal-directed therapy, using all the available hemodynamic data (Scvo2, blood pressure, central venous pressure, and hemoglobin). (n=85). All, of course, also had acute lung injury or ARDS.
They then performed essentially two case/control analyses using early transfusion as the exposure:
- 53 of the 285 septic shock patients got blood within 24 hours; they were compared against the 232 who were not transfused.
- 85 patients "should have" gotten blood early as part of EGDT. The 20 who did were compared to the 65 who did not.
They did a multivariate regression analysis. There was some missing transfusion data, so they used "multiple imputation by chained equations" and "Rubin's rules" to fill in the blanks. (Hmm.) They acknowledge their power was limited (minimum detectable mortality difference 19%).
Of the 285 patients with septic shock (with or without "transfusion criteria"), here's the data at day 28:
- Death: 23 of 53 transfused patients (43%) died, compared to 70 of 232 (30%) non-transfused patients (p=0.06).
- Median ventilator-free days: Zero in the transfused patients, 9 in the non-transfused (p=0.35)
Of the subset of 85 patients with "transfusion-indicated" septic shock:
Only 20 of 85 (24%) got blood in the first 24 hours. They may have died more (50% vs. 29%, p=0.09) and had fewer ventilator free days (0 vs. 9, p=0.26), but there were huge biases as to who got blood, with older, sicker, men more likely to be transfused. It couldn't be identified who got the blood within the first 6 hours, most consistent with EGDT.
After multivariate analysis, long story longer, there was no detectable association between transfusion and mortality or ventilator free days. After inserting or excluding their imputed data, their findings were not demonstrably different.
I'm grateful for this analysis, but thanks to the confounders here, and the small sample size, I still don't know what to conclude or believe. (I guess neither do the authors, who urge caution drawing conclusions and point out the wide confidence intervals traversing benefit and harm. They do use their discussion section to throw a lot of interesting, evidence-based rocks at RBC transfusion after the first six hours in septic shock.)
Parsons EC et al (ARDSNet investigators). Red blood cell transfusion and outcomes in patients with acute lung injury, sepsis and shock. Crit Care 2011; 2011, 15:R221. FREE FULL TEXT