Fresh frozen plasma (FFP) and cryoprecipitate are used widely in hospital practice across a range of clinical specialties, especially in critical care units [1–5]. FFP is prescribed for two main indications: first, to prevent bleeding (prophylaxis), and second, to stop bleeding (therapeutic). However, there are few detailed, prospective, descriptive data from large studies describing these patterns of use in the critically ill. FFP is associated with well-recognised risks [6, 7], some of which are probably underrecognised, such as transfusion-associated lung injury and transfusion-associated circulatory overload [8–13]. Moreover, recent concerns that red cell transfusion may adversely affect clinical outcomes raises concerns that other blood components may not always benefit patients [14–17].

We previously reported a large national study of coagulopathy in UK intensive care units (ICUs), which showed that 30% of all critical care admissions had at least one episode of prothrombin time (PT) prolongation [18]. We described risk factors for the development of coagulopathy and showed an association with higher ICU mortality, but did not describe the use of FFP in detail. The aim of this analysis of the study database was to describe plasma use in detail, with the following research questions. (1) What reasons do clinicians give for administering FFP, and how does the presence of bleeding and coagulopathy influence the dose? (2) What PT or international normalised ratio (PT/INR) triggers are currently used by clinicians, and how do these relate to the doses administered? (3) What changes in PT/INR occur following FFP administration? We also describe the current use of cryoprecipitate in UK critical care, since it is often prescribed with or as an alternative to FFP.

A detailed description of the Intensive Care Study of Coagulopathy (ISOC study) has been previously published [18]. ISOC was a prospective observational cohort study of sequential patients admitted to 29 ICUs in Scotland, England, Northern Ireland and Wales. ICUs were mixed general units and did not include specialist cardiac centres. All admissions were eligible for inclusion, with the exception of readmissions during the same hospital stay and patients with short life expectancy (under 4 hours). The study was observational, and only routinely available clinical data were recorded. The ethics committee waived the need for patient consent (Lothian Research Ethics Committee).

Data were collected prospectively from patient records for the 24 hours prior to admission and every subsequent day in the ICU until discharge, death or 30 days (for patients remaining in the ICU at this time). Data collected included haematology and coagulation test results, all procedures, the occurrence of clinically significant haemorrhage during each 24-hour period (defined as estimated total cumulative blood loss >300 ml, 1 U of red cells or bleeding from a critical site, such as intracranial, as defined in the Audit of Transfusion in Intensive Care in Scotland study [19]), and all blood component transfusions. The timing of FFP transfusion was recorded, and the reasons for FFP were classified by clinical staff into four categories: 'coagulopathy with bleeding', 'coagulopathy with no bleeding', 'coagulopathy with no bleeding prior to an invasive procedure' or 'other'. An additional data collection form was completed on the first day that a patient developed laboratory evidence of coagulopathy (defined below) and included a range of variables predefined by consensus by the investigators as potential factors that might influence the occurrence of coagulopathy.

All data were scanned into a database using the ReadSoft system ('Eyes and Hand', Readsoft Ltd, Milton Keynes, UK), and quality control procedures included computer-based sense checking, manual checking of all queries and resolution of data queries by research staff in individual ICUs when necessary.

The definition of coagulopathy was based on a laboratory threshold INR value >1.5 or an equivalent PT beyond local reference range. Around one-half of participating ICUs reported results as PTs and one-half reported them as INRs. We therefore converted data to an INR value using the local International Specificity Index (ISI) for the laboratory, which enabled comparison across all ICUs.

This pragmatic national epidemiological study of admissions to general ICUs in the UK indicated that 13% of all patients admitted received FFP. There was wide variation in the dose used, even after we classified transfusion episodes according to the reasons given by clinicians, the presence of bleeding and the presence of coagulopathy defined by INR prolongation. A remarkably high proportion of FFP was administered to patients without evidence of bleeding, and many of these patients had normal or only mildly deranged INR values. For example, 40% of all FFP treatments were administered to patients without haemorrhage and in whom preceding coagulation tests were normal or only modestly deranged (INR ≤2.5), and 29% of FFP treatments were administered outside episodes of INR prolongation. Relevant clinical factors to describe the epidemiology of these subgroups of coagulopathy in critical care have been described previously [18], but our data in this paper extend these findings and illustrate the very uncertain rationale underpinning FFP administration, which must raise questions about the appropriateness of use and effectiveness of FFP.

We found that clinicians prescribed marginally higher volumes of FFP in association with bleeding and when the INR was prolonged, and our regression model found haemorrhage to be the clinical indication that was the strongest factor in predicting higher transfusion volume. However, typically only around one additional unit of FFP was administered per transfusion episode under these circumstances, with a median volume transfused of about 11 ml/kg. Current guidelines support the need for adequate doses of FFP to achieve haemostasis, typically 10 to 1 ml/kg [20, 21], but even these doses of FFP may fail to correct low coagulation factor levels without taking into consideration factors such as the size of the patient, the extent of coagulation deficiency or the degree of abnormality to be corrected [2, 22].

Our data suggest that under circumstances where FFP has a higher probability of patient benefit, namely, bleeding in a patient with INR prolongation, inadequate doses are currently administered to reliably correct coagulation factor deficiencies. This may explain, in part, the small decreases in INR prolongation that we observed after FFP transfusion, especially when the INR was only modestly prolonged [2, 23, 24]. Variable reductions in INR may also be explained by differences in the time taken for rechecking INR after FFP transfusion, although our statistical model incorporated this variable. The reasons for transfusion of inadequate volumes of FFP are unclear and require further investigation. In addition, an understanding of the nonlinear relationship between coagulation factor content and INR would also indicate why, with minimally deranged prolongations in INR, the effects of FFP transfusion aimed at raising the levels of coagulation factors and reducing the INR are likely to be very small [25]. If inadequate FFP administration influences clinical outcomes during bleeding, as is suggested by emerging evidence relating to trauma-associated coagulopathy and massive haemorrhage [26, 27], the variation and inconsistent practice we have documented could affect clinical outcomes.

Conversely, the large number of FFP transfusions administered to nonbleeding patients without INR prolongation or with minor derangements suggests the possibility of widespread and inconsistent use of FFP without strong clinical rationale. Recent systematic reviews indicate the poor quality of existing evidence, including the poor predictive value of standard coagulation tests, and no clear benefit of prophylactic use or to reduce bleeding related to planned procedures [28, 29]. Indeed, available data suggest that many patients with isolated minor abnormalities in PT or INR have adequate individual coagulation factor levels, which would be consistent with our data [30]. Furthermore, the FFP volumes administered in these nonbleeding situations were 9 ml/kg or less in more than 50% of cases, and these doses are highly unlikely to result in consistent increases (if any) in coagulation factor levels if deficiencies do exist [2, 22–25]. Our data show that critical care clinicians in the UK continue to use FFP widely in this prophylactic situation, with evidence of insignificant changes in the INR value [2, 31].

Stronger evidence is needed to guide clinicians. Specifically, evidence suggesting that correction of INR prolongation is not required in the prophylactic clinical situation could significantly alter current practice and potentially reduce demand for FFP. Continuing reports of an association between FFP administration and adverse patient outcomes in patients, including the critically ill, further emphasise the requirement to improve the evidence base, especially for transfusion of FFP to nonbleeding patients [6, 9–13, 32–35].

Our population estimates found that 0.63 FFP units/ICU admission were utilised in UK general critical care units, which equates to approximately 63,000 units/year, based on the 100,000 patients admitted to general ICUs in the UK each year. This comprises around 21% of the total FFP used in the UK and, of note, excludes cardiac ICUs, which were not included in our study. On the basis of our estimates of the proportion used in nonbleeding patients with normal or mildly deranged INRs (<2.5), who received 40% of FFP treatments, and our observed median use for this indication of 2 FFP units/transfusion episode, we estimate that about 40,000 units are transfused each year to patients in whom the evidence base for benefit is weak. These estimates illustrate the potential impact of an improved evidence base, particularly if this evidence supports more restrictive use. The potential cost implications of introducing newer products, such as factor complex concentrates or pathogen-inactivated plasma, are also unclear in the absence of evidence of benefit for the standard product.

There are limitations to our pragmatic study. Around one-half of the participating ICUs reported coagulation results as PT and one-half reported them as INRs. The definition of coagulopathy was based on PT prolongation and a threshold INR value >1.5. This cutoff level is commonly used by clinicians and described in the literature [19, 21, 36], and the use of the INR allowed the data to be compared and standardised across the ICUs, which would not be possible for the PT, given that laboratories use different thromboplastins with different laboratory-specific mean normal PT and ISI. However, the PT was developed to assess coagulation factor deficiencies, and the INR was developed to standardise assessment of the response to vitamin K antagonists [36, 37]. For our primary analysis, we did not relate FFP use to APTT, which could also trigger correction of coagulation, especially in relation to bleeding. To explore the potential importance of APTT values, we examined the 56 patients who received 64 FFP treatments but never had documented PT prolongation during their ICU stay, and then extracted the APTT value prior to FFP transfusion. The results did not suggest that APTT prolongation was likely to explain the widespread use of FFP when the INR was normal or mildly prolonged, although our study was not designed to answer this question.

FFP continues to be widely used as a prophylactic or precautionary treatment in critically ill nonbleeding patients with mild coagulation abnormalities. Wide variation in clinical use exists in terms of indication and the dose used, which is often inadequate to reliably increase individual factor levels. Our data support the need for more research to improve the current understanding of the risk-to-benefit profile of plasma therapy, and specifically the need (if at all) for FFP or other products, especially in relation to nonbleeding patients.