Blood transfusion in the ICU (part 4): Transfusion thresholds for FFP and platelets

Read part 1 (Overview), then part 2 (Cross-matching), and finally part 3 (Transfusion thresholds for PRBCs)

Now that we’ve investigated transfusion goals for PRBCs, let’s turn our attention to plasma and platelet products.

Fresh frozen plasma

Transfusion of FFP is indicated to correct coagulopathies. However, clotting factors behave very differently in actively bleeding patients compared to those who aren’t. The former need a reasonably normal clotting profile to achieve hemostasis; the latter can usually tolerate significant coagulopathy without clinical implications.

No single clotting assay captures the entire cascade, but since we are simple creatures, many clinicians like to look at the PT/INR as a global measure of clotting (which it isn’t); from that cloudy perspective, patients who aren’t bleeding or at risk of bleeding can often tolerate extremely high INRs without complications. The most common scenario where this occurs is warfarin toxicity, and in stable patients this can usually be managed simply by stopping warfarin, with or without the addition of vitamin K. In cases where major or life-threatening bleeding is present, or an emergent invasive procedure is planned, more active correction is appropriate… but in the vast majority of cases, it makes more sense to give prothrombin complex concentrates.

Modern 4-factor PCCs (aka KCentra in the US) are mostly safe and—since they carry high concentrations of factors II, VII, IX, and X, the exact factors blocked by warfarin—are far more effective at correcting warfarin anticoagulation than plasma. The only real downside is cost, but when balanced against the risks of transfusing multiple units of plasma, it’s generally a win. After all, blood products convey a significant expense to the system as well.

In non-warfarin-related coagulopathies, giving PCCs is often less effective, since other factors are also absent. Here, transfusing FFP may make more sense—but a rational approach is needed, particularly when addressing minor INR elevations (<1.5–1.7 or so). Further lowering of these levels is difficult to achieve and often fruitless to attempt, as the impact of transfusion becomes less and less as the baseline approaches normal.

There is a medical rumor that FFP itself has an intrinsic INR of 1.7, and hence that transfusion inherently cannot achieve correction below this level. This is incorrect—FFP has the same INR range as normal plasma (1, 2)—but the myth probably originates from data that:

  1. Transfusion below this level tends to cause trivial changes in INR (Abdel-Wahab et al. found that transfusing from a starting INR of 1.0–1.8 only lowered it by ~0.07.)
  2. An INR in this range probably reflects clinically normal clotting times, despite the lab abnormality.
  3. Most INRs in this range tend to correct just as fast with treatment of the underlying disease, even without transfusion.

Holland et al. has the best piece on this topic, where they also derive a useful equation to guesstimate the INR reduction per unit of FFP transfused: 0.37 * [pretransfusion INR] – 0.47.

A few other circumstances:

  • Procedures with a low risk of bleeding (minimally invasive or easily compressible sites), such as most line placements: the Society of Interventional Radiology recommends a routine INR goal of <2–3. In reality, risk from most ultrasound-guided lines placed by experienced operators is probably linked to user skill more than patient coagulopathies; in very ill patients whose numbers cannot be easily corrected, in my humble opinion there is no INR that should preclude necessary line placement by an expert user.
  • Procedures with a higher risk of bleeding (including most abdominal or spinal sites): INR <1.5–1.8, per the Society of Interventional Radiology. As noted, have fun if you attempt to transfuse to reach the lower end of that range.
  • Cirrhotic patients: generally ignore the INR, which does an extremely poor job of capturing true clotting function. Many of these patients have a high INR but may be more likely to pathologically bleed than clot, due to a variety of complex factors not captured by the INR, such as elevated Factor VIII and decreased proteins C/S. A better guide may be the patient’s clinical picture, such as their tendency to ooze from needlesticks. (You can always throw some vitamin K at them if you suspect an element of malnutrition.)
  • Neurosurgery patients: do whatever the neurosurgeons want.
  • Disseminated intravascular coagulopathy: Plasma will not treat the disease, but won’t worsen it either (“feeding the fire”). If bleeding occurs, transfusion is okay.
  • When fibrinogen levels are low: since fibrinogen is a cornerstone of the clotting cascade, this is best corrected in high concentrations with cryoprecipitate, not FFP. Target a level >100–150 mg/dl in serious active bleeding. Usual dose is 5–10 units.

Remember that FFP is a colloid and hence remains largely intravascular. Transfusions of multiple units can therefore result in significant volume loading, perhaps in excess of what you might expect, and in susceptible patients could induce congestive heart failure.


Even more so than clotting factors, platelet requirements are heavily determined by bleeding risk.

  • Actively bleeding patients should usually have a platelet level >50,000
  • Serious bleeding (or bleeding involving the CNS) should have a level >100,000
  • Massive hemorrhage may even benefit from a level >150,000

On the other hand, non-bleeding patients usually tolerate a level of 10–20k or even lower. Where to draw the line is often determined by how hard it is to raise it; very “platelet refractory” patients or those actively destroying platelets (such as in ITP) may barely increase their level no matter what you give them, and you may need to allow them to coast to 10k or lower. However, there is some concern for spontaneous intracranial hemorrhage occurring when patients get below 20k, and especially below 10k.

As for procedural goals:

  • Procedures with a low risk of bleeding (as defined above): >20k. The same caveat noted above applies here; if a patient needs a line and can’t get “in range,” do place the line, just do it carefully and with scrupulous hemostasis.
  • Procedures with a higher risk of bleeding: >50k.

When numbers don’t matter

There are a few cases when following lab values is flat-out the wrong approach to guide transfusion.

For PRBCs: when anemia is obviously symptomatic or contributing to shock.

For platelets: when there is qualitative (not quantitative) dysfunction of the circulating platelets, most commonly due to uremia, or occasionally cirrhosis. Antiplatelet drugs like aspirin can also be “reversed” by introducing healthy platelets, although not in the setting of spontaneous intracerebral hemorrhage, where it actually worsens outcomes. DDAP can also be used to temporarily normalize platelet function and may be an alternative to platelet transfusion.

For massive transfusion: in a rapidly hemorrhaging patient, all lab values lag behind the clinical milieu, and are too slow to draw and process anyway. Instead, transfuse empirically in balanced ratios targeting vital signs. When things start to slow down, you can transition to a lab-based strategy. The one exception might be if viscoelastic testing (TEG or ROTEM) is available.


Transfusing asymptomatic patients to maintain hematology labs within a “normal” range is a common time-killing activity in the ICU, but this decerebrate test/transfuse cycle ignores a much more complex underlying physiology. Play the number game when necessary, but look at the patient as well; in general, the’ll tell you what products they need.