Tag Archives: transfusion

How Quickly Does Hemoglobin Drop After Acute Bleeding?

We all know that hemoglobin / hematocrit drop after blood loss. We can see it decreasing over the days after acute bleeding or a major operative procedure (think orthopedics). And we’ve been told that the hemoglobin value doesn’t drop immediately after acute blood loss.

But is it true? Or is it just dogma?

A reader sent me a request for some hard references to support this. When I read it, I knew I just had to dig into it. This is one of those topics that gets preached as dogma, and I’ve bought into it as well.

Now, I have personally observed both situations. Long ago, I had a patient with a spleen injury who was being monitored in the ICU with frequent vital signs and serial blood draws (but I don’t do that one anymore). He was doing well, then became acutely hypotensive. As he was being whisked off to the OR, his most recent hemoglobin came back at 10, which was little changed from his initial 11.5 and certainly no independent reason to worry.

But hypotension is a hard fail for nonoperative solid organ management. In the OR, anesthesia drew another Hgb at the end of the case, and the value came back 6.

Similarly, we’ve all taken care of patients who have had their pelvis fixed and watched their Hgb levels drop for days. Is this anecdotal or is it real? The doctor / nursing / EMS textbooks usually devote about one sentence to it, but there are no supporting references.

I was only able to locate a few older papers on this. The first looked at the effect of removing two units of red cells acutely. Unfortunately, the authors muddied the waters a little. They were only interested in the effect of the lost red cell mass on cardiac function, so they gave the plasma back. This kind of defeats the purpose, but it was possible to see what happened to Hgb levels over time.

Here were there findings over time for a group of 8 healthy men:

Time Hbg level
Before phlebotomy 14.4
1 week after 11.7
4 weeks after 12.6
8 weeks after 13.6
16 weeks after 13.9

So the nadir Hgb value occurred some time during the first week after the draw and took quite some time to build back up from bone marrow activity.

That’s the longer term picture for hemoglobin decrease and return to normal. What about more acutely? For this, I found a paper from a group in Beijing who was trying to measure the impact of Hgb loss from a 400cc blood donation on EEG patterns. Interesting.

But they did do pre- and post-donation hemoglobin values. They found that the average Hgb decreased from 14.0 to 13.5 g/dl during the study, which appeared to be brief. Unfortunately, this was the best I could find and it was not that helpful.

Bottom line: Your patient has lost whole blood. So, in theory, there should be no Hgb concentration difference at all. But our bodies are smart. The kidneys immediately sense the acute hypovolemia and begin retaining water. The causes ongoing hemodilution within seconds to minutes. Additionally, fluid in the interstitial space begins to move into the vascular space to replace the volume lost. And over a longer period of time, if no additional fluid is given the intracellular water will move out to the interstitium and into the vascular space.

But these things take time. There is an accelerating curve of hemodilution that takes place over hours. The slope of that curve depends on how much blood is lost. A typical 500cc blood transfusion will cause a 0.5 gm/dl drop over several minutes to an hour. We don’t have great data on the exact time to nadir, but my clinical observations support a hyperbolic curve that reaches the lowest Hgb level after about 3 days.

Unlike this curve, it levels off and slowly starts to rise after day 3-4 due to bone marrow activity.

The steepness of the curve depends on the magnitude of the blood loss. After a one unit donation, you may see a 0.5 gm/dl drop acutely, and a nadir of 1 gm/dl. In the case of the acutely bleeding patient with the spleen injury, the initial drop was 1.5 gm/dl. But two hours later it had dropped by over 5 gm/dl. 

Unfortunately, the supporting papers are weak because apparently no one was interesting in proving or disproving this. They were more interested in cardiac function or brain waves. But it does happen. 

Here’s the takeaway rule:

In a patient with acute bleeding, the initial hemoglobin drop is just the tip of the iceberg. Assume that this is only a third (or less) of how low it is going to go. If it has fallen outside of the “normal” range, call for blood. You’ll need it!


  1. Effect on cardiovascular function and iron metabolism of the acute removal of 2 units of red cells. Transfusion 34(7):573-577, 1994.
  2. The Impact of a Regular Blood Donation on the Hematology
    and EEG of Healthy Young Male Blood Donors. Brain Topography 25:116-123, 2012.


Best of EAST #2: Blood Transfusion And Nosocomial Infection

This abstract falls into the “interesting, but how can we use this bit of information” category. We’ve known that transfusing packed red cells raises nosocomial infection rates for at least 15 years. The group led by MetroHealth in Cleveland combined forces with the Vanderbilt trauma group to re-look at their data from the PAMPer trial with respect to trauma patients.

The PAMPer trial (Prehospital Air Medical Plasma) examined the effect of tranfusion of two units of plasma in the air ambulance on mortality, transfusion need, and complications. Half of the patients got plasma plus standard care, and the other half standard care alone.

This abstract uses PAMPer trial data to examine the impact of giving any amount of blood on nosocomial infection in these patients. These infections included pneumonia, bloodstream infection, C Diff colitis, empyema, and complex intra-abdominal infection.

The group retrospectively analyzed the prospectively collected PAMPer data and used logistic regression models to test for an association.

Here are the factoids:

  • A total of 399 patients with the usual trauma demographics were included (younger male, moderately injured, blunt mechanism)
  • Ten percent of patients died, and 23% developed nosocomial infections
  • Pneumonia was by far the most common complication (n=67) with all others in the low teens or below
  • Although only two thirds of patients received plasma, 80% were given PRBCs and 27% received platelets
  • Patients who received any amount of packed cells had a 2.3x increase in nosocomial infections, and the number given increased the rate of nosocomial infection (1.06x)

The authors concluded that patients in the PAMPer trial who received at least one unit of blood “incurred a two-fold increased risk of nosocomial infection” and that this risk was dose dependent.

My analysis: The biggest obstacle for me to buy into this work is the enrolled patient group. Studies in which you borrow someone else’s data are always a bit problematic. You have no control over the variables, as they’ve been determined by someone else.

In this case, the dataset could only be controlled for age, sex, and ISS. But what about all the other stuff that might have an impact on infections? Things like pulmonary injury, the 20% of patients who had penetrating injury, and severe TBI patients with their propensity to develop VAP.

The odds ratios of the associations were a bit on the low side. Sure, the overall nosocomial infection odds ratio was 2.37 but the confidence interval was 1.14 to 4.94. This is very wide and it means that the odds could have been anywhere from 1.14x to almost 5x. This suggests that the study group may not have been large enough to give us a clear picture. And the odds ratio for number of PRBC units vs infection was only 1.06 with a tighter confidence interval. So even if it is present, this association is very, very weak. I like to see ridiculously large odds ratios when reviewing observational studies like this where the input data is constrained.

My final comment on this study deals with its utility. These are trauma patients. They are bleeding. We’ve known that transfusions may increase the nosocomial infection rate in critically ill patients for at least 15 years. But we will still have to give the patients blood. So what are we to do?

Here are some questions for the authors and presenter:

  • Please comment on the limitations you faced using the PAMPer dataset. Were you satisfied with the range of variables available? Which additional ones would you have liked to work with?
  • Do you feel that the 399 patients provided enough statistical power for analysis? The confidence intervals are large and very close to the OR=1 line.
  • What should we do with your conclusions? Can we translate this into clinical practice?

One final note: the patients did not “incur increased risk.” Rather, there was an association with increased risk of infection. We really don’t know if it was from the blood or something else that was not recorded in the PAMPer dataset.

Reference: Dose-dependent association between blood transfusion and nosocomial infections in trauma patients: a secondary analysis of patients from the PAMPer trial. EAST 2021, Paper 3.

Can We Use Type A Plasma For Emergency Transfusion?

Trauma patients tend to try to bleed to death. And trauma professionals try to stop that bleeding. They also frequently have to replace the blood products that were lost, which includes red blood cells, plasma, platelets, and more.

From a red blood cell standpoint, we have a long history of using group O- packed red cells as the so-called universal donor product. The problem is that only about 5% of the world population has this blood type, so it can be scarce.

To address this, many centers have moved toward using O+ blood for select patients. This blood type is much more prevalent (about 50% worldwide). The only difference is the positive Rh factor which has little impact on males, or females who are not in their child-bearing years. If an allergic reaction occurs, it is typically mild.

But what about plasma? This is interesting stuff. When selecting red cells, we want them to have no ABO group antigens on them so they don’t provoke a reaction. But plasma is just the opposite. We don’t want any ABO group antibodies in it. And the only plasma without antibodies comes from people who have all of them (A and B) on their red cells. This means people with type AB+ blood. Unfortunately, this is the other rare blood type, so there’s not a lot to go around. Worldwide, about 5% of people are AB+ and less than 1% are AB-.

So why couldn’t we do something like we did with packed red cells and substitute a more common blood type that evokes little immune response? The American Association of Blood Banks (AABB) has authorized both AB and A plasma for use in emergency situations. Unfortunately, the safety profile for using group A has not been very well studied, particularly in trauma patients needing massive transfusion.

The authors of the PROPPR study re-analyzed the data from it to try to answer this question. As you may recall, PROPPR was published in 2015 and compared safety and effectiveness of transfusion ratios at 1:1:1 to 1:1:2 (plasma : platelets : red cells).

The study group selected patients from the dataset who received at least one unit of emergency release plasma (ERP), defined as product given before the patient’s ABO type had been determined. Nicely enough, 12 sites transfused group AB ERP and 9 sites gave group A. One site gave both A and AB.

The authors looked at in-hospital mortality at 30 days, and a host of complications. Here are the factoids:

  • A total of 584 of the 680 patients in the PROPPR study received emergency release plasma
  • The median number of units given was 4, and there was no difference between A and AB groups
  • There were statistically significant baseline differences between the groups, including blood type, SBP, percent in shock (SBP<90), blunt mechanism, positive FAST that were probably not very clinically significant
  • The number of transfusions of all products were significantly  higher in the A plasma group
  • Complications were significantly higher in the A plasma group, specifically from SIRS, pulmonary problems, and venous thromboembolism (VTE)
  • There were no acute hemolytic transfusion reactions and three febrile reactions

The authors concluded that, statistically, the use of group A plasma was not inferior to the use of group AB. The authors stated that cautious use of group A is an acceptable option, especially if group AB is not readily available.

Bottom line: Here we go again. Always be careful when reading a study that suggests non-inferiority of one thing compared to another. There are a lot of potential issues here:

  • The PROPPR trial data was not designed to answer questions about plasma usage, so the data is being highjacked a bit
  • Participating centers did not have a standardized way to determine the group that received ERP, so some data anomalies will be present
  • The A and AB study groups were different in many ways at baseline, particularly with respect to how much product they received
  • The primary outcome, 30-day mortality, was underpowered and could never show a significant difference

So with significant baseline differences in study groups and a potentially underpowered study, don’t read non-inferiority as meaning that use of group A plasma is okay. We still just don’t know. What this study really shows is that you can “get away with” using low titer group A plasma if you run out of AB. But it shouldn’t be your go to product yet. To figure out the real safety profile, we need to do a real “PROPPR” study. Get it?

Reference: Group A emergency-release plasma in trauma patients requiring massive transfusion, J Trauma 89(6):1961-1067, 2020.

Best Of EAST 2020 #2: Do Platelet Transfusions Fix Sad Platelets?

The next abstract from EAST tackles the question of how we can treat platelets that don’t work right in trauma patients. The literature on using platelet transfusions in patients who are taking anti-platelet agents is getting fairly clear: they don’t work. But what about for platelets that don’t work right due to traumatic hemorrhage?

The trauma group at Penn attacked this problem by performing a prospective study at their Level I trauma center. They investigated platelet function using thromboelastography (TEG) with platelet mapping on trauma patients admitted to the intensive care unit over a two year period. They analyzed platelet function and counts at 3, 6, 9, 12, and 24 hours after admission. Platelet function in patients given platelets during any of the intervals were compared to those who were not. Outcomes studied were improvement in platelet function and mortality.

Here are the factoids:

  • A total of 93 patients were entered into the study
  • About half (57%) had platelet dysfunction detected by TEG
  • Mortality was not different between the groups
  • Neither platelet count nor function improved with transfusion

The authors concluded that platelet dysfunction is common in these patients and that platelet transfusions do not appear to restore platelet function.

My comment: This abstract is a bit hard to follow. Hopefully the manuscript will have more detailed tables that break down which patients got platelets and at what times. It appears that patients could have gotten platelets at various times (any, all, or none) after admission to the ICU, and that pre- and post-transfusion TEG runs were analyzedfor each. It’s also not clear if every patient with dysfunction got a transfusion.

The most obvious issue here is that the total number of patients is small, and the numbers getting platelets at each time interval is even smaller (10-49). The statistical power of such a study is very low. It’s not surprising that no significant differences could be detected. This means that failing to see significance doesn’t means it’s not necessarily there, just that many more patients are needed. So it’s hard to buy into the assertion that platelet transfusions don’t matter.

Here are my questions/comments for the presenter:

  1. Why didn’t all patients get platelets? From the table, it looks like nearly all patients had significant dysfunction (defined as MAadp < 40mm) until the end of the 24 hour study period. It looks like some selection bias is possible if there was no defined protocol for giving transfusions to those who had an abnormal TEG.
  2. Is your study sufficiently powered to draw the conclusion it did? The number of patients seems small overall, and doing measurements serially every 3 hours would seem to further weaken the statistics. Please comment on your choice of analysis and how likely you are to actually be able to detect significance.
  3. Be sure to clarify the details of when platelets were given and why, how many measurements were taken and when, and exact patient numbers. These are not clear in the abstract due to length limitations.

This paper is very interesting and I look forward to its presentation.

Reference: Platelet infusions do not correct trauma induced platelet dysfunction. EAST Annual Assembly abstract #24, 2020.

Massive Transfusion: What’s The Right Ratio?

In my last post, I analyzed a survey that studied the massive transfusion protocol (MTP) practices of academic Level I trauma centers in the US. What centers do is one thing. But what does the literature actually support? A group from Monash University in Melbourne, Australia and the National Health Service in the UK teamed up to review the literature available through 2016 regarding optimal dose, timing, and ratio of products given during MTP.

One would think that this was easy. However, the search for high quality ran into the usual roadblock: the fact that there is not very much of it. The authors scanned MEDLINE for randomized, controlled studies on this topic, and found very few of them. Out of 131 articles that were eligible, only 16 were found to be suitable for inclusion, and 10 of them were still in progress. And only three specifically dealt with the ratio question. Even they  were difficult to compare in a strict apples to apples fashion.

Here are the factoids that could be gleaned from them:

  • There was no difference in 24-hour or 30-day mortality between a ratio of 1:1:1 (FFP:platelets:RBC) vs 1:1:2
  • However, a significantly higher number of patients  achieved hemostasis in the 1:1:1 group (86% vs 78%)
  • There was no difference in morbidity or transfusion reactions in the two groups
  • One study compared 1:1 component therapy with whole blood transfusion and found no difference in short-term or long-term mortality or morbidity

Bottom line: As usual, the quality of available data is poor if one limits the field to randomized, controlled studies. Ratios of 1:1:1 and 1:1:2 appear to be equally effective given the limited information available. A number of papers not included in this review (because of their less rigorous design) do seem to indicate that higher ratios of RBC (1:3-4) appear to be detrimental. And as time passes, more and hopefully better studies will be published.

What does this all mean for your MTP? Basically, we still don’t know the best ratio. However, it is recommended that your final ratios of FFP:RBC end up somewhere between 1:1 and 1:2. The only way to ensure this is to set up your MTP coolers so the the ratio of product they contain is better than 1:2. This means more plasma than 1 unit per 2 units of red cells. 

If you set it at the outside limit of 1:2, then that is the best ratio you can ever get assuming everything goes perfectly. However, if you have to thaw frozen plasma, use too much emergency release PRBC before activating MTP, or someone cherry-picks the coolers to transfuse what they think the patient needs, the ratios will quickly exceed this boundary.

So be sure to load your coolers with ratios that are closer to 1:1 to ensure that your final ratios once MTP is complete are what you want them to be. And monitor the final numbers of every one of your MTP activations through your trauma performance improvement program so you know what your patients are really receiving.

Reference: Optimal Dose, Timing and Ratio of Blood Products in Massive
Transfusion: Results from a Systematic Review. Transfusion Med Reviews 32:6-15, 2018.