SEARCH

SEARCH BY CITATION

Keywords:

  • Blood transfusion;
  • Fresh frozen plasma;
  • Randomised controlled trials

Summary

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

Fresh frozen plasma is commonly used in cardiac surgery in an attempt to replace clotting factors and to decrease bleeding. Despite this, there has been no previous review of the available literature to support this practice. The aim of this review was to study the effect of prophylactic peri-operative transfusion of fresh frozen plasma on bleeding and coagulopathy in patients undergoing cardiac surgery. A comprehensive literature search was performed and all randomised controlled trials of the use of fresh frozen plasma in cardiac surgery were included. Six small trials were found that included a total of 363 participants with six different dose regimens of fresh frozen plasma. The overall quality of the studies was poor due to small patient numbers and lack of allocation concealment. There was no evidence that the prophylactic use of fresh frozen plasma affected peri-operative blood loss in cardiac surgery. There was some evidence that it may improve platelet count and fibrinogen concentration.

Fresh frozen plasma (FFP) is frequently used after cardiac surgery to reduce excessive postoperative bleeding. Although UK guidelines clearly state that FFP should only be used if there is evidence of abnormal coagulation [1], audit data has shown that FFP is often given inappropriately [2,3]. A study of five hospitals in the South of England found that 23% of the hospitals' FFP was used in cardiothoracic and thoracic surgery [4]. Very few controlled trials exist that evaluate the benefit of such practice in terms of decreasing blood loss and correcting coagulation abnormalities. The administration of FFP does no harm to the majority of recipients. However, there is some risk of transfusion-related complications and infections. Fresh frozen plasma is implicated in at least 30% of the cases of transfusion-related acute lung injury reported in the UK each year (2001–2002 Serious Hazards of Transfusion (SHOT) report [5]). The Handbook of Transfusion Medicine [6] states that although FFP does not decrease transfusion requirements, it does help to correct prolonged clotting times and to improve haemostasis. What is less clear is the benefit to the patient in terms of blood loss reduction in the cardiac bypass setting. This review summarises the results of six randomised, controlled trials that examine the efficacy of peri-operative transfusion of FFP in decreasing blood loss irrespective of coagulation results. The ability of FFP to correct coagulopathy is also assessed.

Methods

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

The search strategy aimed to find randomised, controlled trials that examined the efficacy of the administration of FFP before, during or after cardiac surgery. Electronic searches of the Cochrane library and Pubmed (1963 – present) covered the period from January 1966 to August 2003. The search strategy involved identifying a combination of the following text terms in abstracts of papers: fresh frozen plasma, FFP, coronary artery, bypass and graft. The resultant studies were then narrowed down, by searching for text terms identifying the studies as randomised, controlled trials. Printed copies of any papers reporting trials with FFP as the intervention were obtained and the reference lists were scanned for further studies. Co-authors were invited to identify any other unpublished trials or those not found in electronic searches. Hand searching of blood transfusion journals was also performed. All randomised, controlled trials of the infusion of FFP in adults undergoing elective cardiac surgery were reviewed. Only trials with a concurrent control of a ‘no infusion’ or ‘placebo infusion’ group were included.

The quality of the trials was assessed for selection bias, i.e. the adequacy of randomisation and allocation concealment, performance and detection bias, i.e. blinding of allocation to patients, carers and investigators, attrition bias, i.e. whether there was postrandomisation exclusion from analysis, and any other forms of bias in the design of the study. The generalisability of the results and the adequacy of reporting of outcomes were also evaluated.

Baseline characteristics of age, sex, weight and previous surgery were evaluated for comparability across treatment arms and data on the most commonly reported endpoints of blood loss, platelet count, fibrinogen levels, haemoglobin concentration and prothrombin time were extracted where possible. Other endpoints were not usable because they were either exclusive to one study or were not adequately reported.

Results are presented as mean (SD). The data were analysed using the standardised mean difference in the Cochrane RevManTMpackage (http://www.cochrane.org/software/download.htm). Sources of heterogeneity were investigated, including differences in baseline characteristics, study fluid volume and quality of study design.

Forest plots

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

Forest plots are used to illustrate the results. A forest plot is a graphical method of summarizing the results of a meta-analysis. Lewis and Clarke provide a good overview in the British Medical Journal [7]: ‘In a typical forest plot, the results of component studies are shown as squares centred on the point estimate of the result of each study. A horizontal line runs through the square to show its confidence interval – usually, but not always, a 95% confidence interval. The overall estimate from the meta-analysis and its confidence interval are put at the bottom, represented as a diamond. The centre of the diamond represents the pooled point estimate, and its horizontal tips represent the confidence interval. Significance is achieved at the set level if the diamond is clear of the line of no effect. The plot allows readers to see the information from the individual studies that went into the meta-analysis at a glance. It provides a simple visual representation of the amount of variation between the results of the studies, as well as an estimate of the overall result of all the studies together’. We have used Metaview 4.2 software to create our forest plots. This format incorporates a table with the plot, which details the individual study, patient numbers and outcome summaries for each treatment group, the weight given to each study, which is proportional to the size of the study, and the effect size estimate of the difference with confidence intervals, corresponding to the point estimates in the plot. Our summary statistic is the standardised mean difference, which is the difference between means divided by the standard deviation. We used a fixed effects model, which assumes all the studies included are estimating one true effect, and that any variation is due to random error. The effect size is plotted on the x-axis, which shows the scale of the differences and is labelled to indicate which side of the vertical line favours control and which favours intervention. The vertical axis simply provides a marker for each study, which may be plotted in alphabetical order of authors or chronological by date of publication.

For example, Fig. 1 shows the mean blood loss after coronary surgery, with data taken from the six trials. The size of each square is proportional to the size of each study. In the first study, slightly less blood is lost in the FFP group than in the control group; this gives a standardised mean difference of −0.11, so the square representing the point estimate is to the left of the line of no effect, favouring FFP. However, the 95% confidence interval for the difference includes zero, so this crosses the vertical line and the difference is therefore not statistically significant. In Boldt's study, there is more blood lost in the FFP group, and the square is to the right of the line, favouring control, again with the confidence interval crossing the zero difference line. The diamond at the bottom is the pooled effect; the centre of the diamond is very near zero, showing no overall effect. A test for heterogeneity is also given; if this is significant, then there is some evidence that the studies do not represent the same patient population. We performed a fixed effects analysis, which assumed there is no heterogeneity in effect sizes. For our blood loss comparison, it appears there may be some heterogeneity; the Kasper results are inconsistent with the others. In such cases, a random effects model can be used, which assumes that the true study-specific effect sizes are not constant but vary randomly about an overall effect. This assumption widens the 95% confidence interval around the estimate; in our blood loss comparison it is widened from −0.22–0.20 to −0.40–0.39. There is also an overall significance test of the effect, which is significant if the pooled estimate is clear of the line of no effect.

image

Figure 1. Forest plot for blood loss results. See text for explanation. Values are blood loss in ml.

Download figure to PowerPoint

Results

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

The search of the Cochrane database found three potential studies, Boldt [8], Consten [9] and Martinowitz [10]. Further searches with Pubmed identified Kasper [11] and Wilhelmi [12] and one study, Trimble [13], was found by hand searching. No unpublished data were found, giving a total of six studies. Scanning the reference lists identified no further studies, and none were identified from other sources. These studies looked at six different doses of FFP, and compared routine administration to either no prophylactic treatment with FFP, or placebo control using hydroxyethyl starch, Gelofusine or packed red blood cells. The six studies are summarised below.

Trimble 1964 (California) [13]

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

This was a randomised controlled trial of 53 patients undergoing cardiopulmonary bypass. Patients with a variety of congenital and acquired heart defects were included. A list of random numbers was used to randomise consecutive patients to receive either two 250 ml units (one unit for children) of FFP (n = 21) after the end of bypass and heparin neutralisation, or not (n = 32). Surgeons were blinded to the randomisation. Chest drainage was recorded at 07:00 of the first postoperative day (between 12 h and 18 h after surgery) and at 24 h intervals. Results were presented separately for adults and children.

Boldt 1989 (Germany) [8]

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

This was a randomised, controlled trial of 40 patients who underwent elective coronary artery bypass graft surgery. No other inclusion criteria were stated. Exclusion criteria were haemorrhagic diathesis, pre-operative treatment (timeframe undefined) with any substance known to affect coagulation, or abnormalities in the pre-operative coagulation screen. Patients were randomised by an unspecified method to receive either two units (430 ml) of prophylactic FFP (n = 20) or not (n = 20) immediately after completion of cardiopulmonary bypass. There was no mention of blinding of patients, carers or investigators. The following outcome measurements were taken before anaesthesia, 30 min after the beginning of bypass, at 5 min, 60 min and 5 h after CPB, and on the first postoperative day: blood drainage volume, haemoglobin concentration, haematocrit, red cell, platelet and leucocyte counts, free haemoglobin, activated partial prothrombin time, prothrombin time, elastase levels and fibrinogen levels.

Martinowitz 1990 (Israel) [10]

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

This was a randomised, controlled pseudo-crossover trial of 40 patients undergoing cardiac surgery with cardiopulmonary bypass. Twenty-six patients were undergoing coronary artery bypass grafting, 11 patients had valve replacements, two had mitral commissurotomy and one had a procedure to close an atrial septal defect. No other inclusion criteria were reported. Patients who had been receiving dipyridamole or aspirin-containing drugs during the two weeks before surgery were excluded. Patients donated fresh whole blood on the morning of surgery, and this was stored at room temperature and processed into fresh plasma and packed red cells. Patients were randomised, by an unspecified method, to receive an unspecified volume of either autologous fresh plasma (n = 20) or autologous packed red cells (n = 20) at the end of cardiopulmonary bypass. There was no mention of blinding of patients, carers or investigators. The following outcome measurements were made before the operation and after infusion of the study fluid: Ivy bleeding time (the time taken for a controlled skin incision to stop bleeding), platelet count, platelet aggregation in response to collagen, adenosine diphosphate, epinephrine and ristocetin levels. Each group later received the other alternate study fluid (pseudo-crossover design), and the above measurements were repeated. The volume of blood loss and blood transfusion requirement was also measured at 24 h after surgery.

Consten 1996 (Amsterdam) [9]

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

This was a randomised, placebo-controlled trial of 50 patients undergoing coronary artery bypass grafting. No other inclusion criteria were stated. Exclusion criteria were pre-existing coagulopathy, a pre-operative left ventricular ejection fraction <40% and cardiopulmonary bypass duration >2 h. Patients were randomised after surgery by an unspecified method to receive either three units of FFP (n = 24) or Gelofusine 750 ml (n = 26) at the end of the operation, using opaque envelopes to conceal the sequence. No FFP was given before or during the operation. All patients were given homologous packed red cells if haematocrit levels decreased to <0.2 after bypass, or <0.25 in the intensive care unit. Double blinding was maintained by having one investigator administer the allocated treatment and a second independent investigator blinded to allocation in the intensive care unit and on the ward. The following outcome measures were evaluated before anaesthesia, after protamine was given, 2 h, 6 h, 24 h and 5 days after surgery: haemoglobin concentration, haematocrit, platelet and leucocyte counts, activated partial thromboplastin time, prothrombin time, fibrinogen level, blood loss and transfusion requirements.

Wilhelmi 2000 (Germany) [12]

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

This was a randomised, placebo controlled trial of 120 patients undergoing coronary artery bypass grafting. Patients undergoing either elective or emergency surgery were included. Exclusion criteria were previous thoracic surgical intervention, oral medication affecting platelet aggregation within 72 h of surgery, a history of coagulation disorders and left ventricular ejection fraction < 40%. Patients were block randomised after completion of cardiopulmonary bypass to receive either four units of FFP (n = 60) or hydroxyethyl starch solution 1000 ml (n = 60). All FFP patients were operated on before controls; consequently there was no blinding of allocation to investigators or carers. At 1 h after surgery, any patient whose cumulative blood loss exceeded a specified rate was given four units of FFP (1000 ml). Packed red blood cells were given only if haemoglobin level decreased to <8.0 g.dl−1. The following outcome measurements were taken before surgery and at 3 h, 6 h, 12 h, 24 h and 48 h after surgery, and also before discharge: haemoglobin concentration, haematocrit, platelet count, prothrombin time and activated partial thromboplastin time. Cumulative blood loss was measured at 1 h, 4 h and 24 h after surgery. The volumes of re-transfused cell saver blood and other transfused blood products were recorded. Myocardial events were documented and economic costs evaluated.

Kasper 2001 (Germany) [11]

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

This was a randomised, placebo-controlled trial of 60 patients undergoing elective coronary artery bypass grafting. Patients on the hospital waiting list for elective surgery were included if they had an anticipated waiting period of four weeks or more, weighed 50–100 kg and met selection criteria for autologous blood donors. Exclusion criteria included left ventricular ejection fraction <40%, haemoglobin <12.0 g.dl−1, total plasma protein <60 g.l−1, medication with drugs that affect blood coagulation taken within seven days of surgery, abnormal coagulation and known allergy to hydroxyethyl starch. An independent individual used a computer-generated sequence to allocate patients randomly before surgery. Thirty patients were allocated to receive autologous FFP 15 g.kg−1, collected during two hospital visits before surgery, and 30 patients were given hydroxyethyl starch 15 g.kg−1. Four patients were excluded from analysis because operative complications rendered the infusion of a large volume of fluid inappropriate. Administration of study fluid occurred at the end of cardiopulmonary bypass, when heparin reversal had been achieved. Patients allocated to receive hydroxyethyl starch did not donate autologous plasma. The investigators were not blinded to the allocated study fluid. In the intensive care unit after surgery, packed red cells were transfused to any patient whose haemoglobin was <9.0 g.dl−1, and FFP and platelets were given to any patient who bled at more than a specified rate and had abnormal coagulation. The following outcome measures were recorded during anaesthesia, after infusion of the study fluid, and at 24 h and 3 and 7 days after surgery: haemoglobin concentration, platelet count prothrombin time, activated partial thromboplastin time, fibrinogen, antithrombin III and albumin. Chest tube drainage volume was recorded for 24 h, and haematocrit was measured 6 h after surgery.

The methodology of the trials varied greatly (Table 1). All studies were small, ranging from 40 to 120 participants. None specified the effect difference they were hoping to show, but each study was inadequately powered to detect all but very large differences in outcome. Three studies were placebo-controlled, one gave the method of randomisation as being by computer, one used a random number list and one mentioned block randomisation. Concealment of allocation was mentioned in one study and was achieved by way of opaque envelopes. Double blinding of allocation to patient and investigator was maintained in the Consten study. In the Trimble trial, blinding was maintained amongst surgeons and theatre staff until administration of the study fluid. Two studies were unblinded, and one did not provide enough information to determine the degree of blinding. No study explicitly stated that the person responsible for measuring outcomes was blinded to allocation. There were multiple outcomes in four studies, that were not accounted for in the analysis. None of the studies stated a clear primary outcome, although three identified one or two main outcomes as being important in the introductory section. One study carried out a per-protocol analysis, with patients being excluded as a result of operative complications, while the remaining studies included all patients but did not mention a planned intention to treat analysis. Two studies used FFP to treat excessive postoperative bleeding before the end of the study regardless of original allocation, which may have caused a dilution of treatment effect. The inclusion and exclusion criteria were similar to those used in common practice. Therefore, the patients selected would be representative of the population of patients needing coronary artery bypass surgery. There were more men than women in the studies this reflects figures from the British Heart Foundation that state that 23% of premature deaths in men and 14% of premature deaths in women are from coronary artery disease [14]. The risk of bias was deemed high if no concealment of allocation or blinding was attempted, and if FFP was given as a treatment for bleeding before assessments were made (Martinowitz, Wilhelmi and Kasper). Moderate risk was determined if there was no concealment of allocation or blinding, and no administration of FFP in recovery (Trimble and Boldt). Studies were classed as low risk if there was some attempt to conceal the allocation, some degree of blinding, and no administration of FFP in recovery (Consten).

Table 1.  Methodological quality of the included studies.
 nPlaceboAllocation concealedInvestigator blindedCarers blindedNumbers analysedFresh frozen plasma in the intensive case unitRisk of bias
Trimble (1964) [13]40YesNot clearInitiallyInitiallyPer protocolNot clearModerate
Boldt (1989) [8]40NoNot clearNot clearNot clearIntention to treatNoModerate
Martinowitz (1990) [10]40NoNot clearNot clearNot clearIntention to treatYesHigh
Consten (1996) [9]50YesEnvelopesYesNot clearIntention to treatNoLow
Wilhelmi (2000) [12]120YesNot clearNoNoIntention to treatYesHigh
Kasper (2001) [11]60YesNot clearNoNot clearPer protocolYesHigh

The mean age of adult patients was given in all studies, and ranged from 42 to 65 years (Table 2). Most participants were male; the male: female ratio was given in four studies and the percentage of male subjects ranged from 58 to 89%. Mean weight was recorded in three studies, and ranged from 74 to 81 kg. The proportion of patients who had undergone previous coronary artery surgery was recorded in Martinowitz as 15%.

Table 2.  Characteristics of the study populations.
StudyInterventionMean age; yearsPercentage of male patientsMean weight; kg
Trimble (1964) [13]Fresh frozen plasma 500 ml42No detailNo detail
None43No detailNo detail
Boldt (1989) [8]Fresh frozen plasma 2 units58No detail79
None62No detail76
Martinowitz (1990) [10]Autologous fresh plasma5975No detail
Autologous packed red cells6270No detail
Consten (1996) [9]Fresh frozen plasma 3 units628381
Gelofusine 750 ml635876
Wilhelmi (2001) [12]Fresh frozen plasma 4 units6372No detail
Hydroxyethyl starch 1000 ml6573No detail
Kasper (2001) [11]Fresh frozen plasma 15 g.kg−1588977
Hydroxyethyl starch 15 g.kg−1568674

The outcomes below were extracted and summarised using the standardised mean difference, which is the difference in means between FFP and control arms divided by the pooled standard deviation. Fixed effects were assumed.

Blood loss at 24 h (Fig. 1)

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

Although the volume of blood loss was mentioned in the Methods section of all the studies, it was only reported at 24 h in three studies. Trimble reported blood loss at 07:00 on the first postoperative day, and Consten reported blood loss for the entire intensive care unit period, neither of which are fixed time frames. The range of total blood loss by 24 h, where reported, was 75–3700 ml. The mean (SD) cumulative blood loss at 24 h was given by three studies: Boldt, Martinowitz and Wilhelmi. Boldt reported a mean (SD) blood loss of 602 (180) ml in patients given FFP and 547 (113) ml in the controls. In the Martinowitz trial, mean (SD) blood loss was 820 (161) ml in the FFP arm and 730 (241) ml in the controls. The mean (SD) blood loss of patients given FFP in the Wilhelmi study was 588 (224) ml compared to 576 (272) ml, to give a combined standardised mean difference between the FFP and control arm of 0.18 (95% CI: −0.09–0.46). To make use of the inadequately reported data on blood loss in the remaining three studies, some data manipulation was used to derive means and standard deviations. Kasper presented the median blood loss at 24 h and, assuming a normal distribution, the mean was estimated as being equal to the median. The SD was estimated from the mean of the available SDs from other studies. This gave a mean (SD) blood loss of 630 (188) ml in the FFP group and 830 (209) ml in controls. The values given in Trimble were estimated to have been measured over approximately 16 h (if the operations began at 07:00 and took approximately 8 h). Therefore, the values were extrapolated to 24 h, and the SDs were calculated as for Kasper. This resulted in a mean (SD) blood loss of 1560 (188) ml in the FFP arm and 1582 (209) ml in the control arm. Consten's data were also included, assuming the intensive care stay duration was approximately 24 h; they reported a mean (SD) blood loss of 896 (509) ml in the FFP group and 776 (388) ml in controls. Including data from all six studies, no overall difference in the volume of blood loss was found; the combined standardised mean difference was −0.01 (95% CI: −0.22–0.20). The largest difference reported by any of the trials was 200 ml more blood loss in the control arm of the Kasper study.

Platelet count (Fig. 2)

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References
image

Figure 2. Forest plot for platelet count results. Values are platelet count × 109.l−1.

Download figure to PowerPoint

The mean platelet count at 24 h was reported by four studies. Boldt showed slightly higher counts in the control group (standardised mean difference = −0.25(85% CI: −0.88–0.37), and Kasper and Consten showed FFP transfusion did not affect platelet count (standardised mean difference = −0.02 (95% CI: −0.54–0.51) and 0.02 (95% CI: −0.53–0.58), respectively). Wilhelmi showed a statistically significant higher platelet count in those given FFP (standardised mean difference = 0.63 (95% CI: 0.26–0.99)). The large numbers in this study gave increased weight to the pooled standardised mean difference of 0.24, which was not quite significant in favour of FFP (95% CI: −0.01–0.48); the confidence interval excluded the possibility that the control was better.

Fibrinogen (Fig. 3)

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References
image

Figure 3. Forest plot for fibrinogen results. Values are fibrinogen level in g.l−1.

Download figure to PowerPoint

Fibrinogen concentration was reported by two studies. The concentration was higher in those given FFP, although all reported mean values fell within the normal range of 2.0–4.0 g.l−1. The standardised mean difference in the Boldt study was 0.27 (95% CI: -0.36–0.89) and was 0.62 (95% CI: 0.08–1.15) in Kasper. This gave a pooled standardised mean difference of 0.47 (95% CI: 0.06–0.87), indicating that the concentration was significantly lower in those in the control arms, although still within the normal range.

Haemoglobin (Fig. 4)

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References
image

Figure 4. Forest plot for haemoglobin concentration results. Values are haemoglobin concentration in g.dl−1.

Download figure to PowerPoint

Haemoglobin concentration was adequately reported in the studies by Boldt and Kasper. Consten did report haemoglobin at 24 h after surgery in the form of a graph, from which only the means could be obtained; the SDs were estimated by pooling those estimated by Boldt and Kasper. Haemoglobin concentrations at 24 h after surgery were low in all groups, Kasper reported a mean value of 10.1 g.dl−1 in both treatment arms, Boldt recorded 12.2–12.5 g.dl−1 and Consten reported values of 12.0–12.1 g.dl−1. The standardised mean difference between the FFP intervention group and control group was −0.25 (95% CI: −0.88–0.37) in the Boldt study, suggesting that haemoglobin concentrations were slightly lower in patients given FFP. Standardised mean differences were 0.00 (95% CI: −0.52–0.52) in Kasper and 0.04 (95% CI: −0.52–0.59) in the Consten study. Overall, there was no evidence of a difference in haemoglobin concentration, with a pooled standardised mean difference of -0.06 (95% CI: −0.38–0.27).

Prothrombin time (Fig. 5)

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References
image

Figure 5. Forest plot for prothrombin time results. Values are prothrombin time expressed as a percentage of normal.

Download figure to PowerPoint

This was reported by Consten at 24 h after surgery in seconds, with a mean (SD) prothrombin time of 20.0 (1.0) in patients given FFP and 17.0 (1.0) s in those given Gelofusine. The remaining studies presented prothrombin time as a percentage of the normal value. In the Boldt trial, patients given FFP had a shorter prothrombin time than those not given FFP, standardised mean difference = 0.14 (95% CI: −0.48–0.76). The administration of FFP also resulted in a shorter prothrombin time in the Kasper study, standardised mean difference = 0.51 (95% CI: −0.03–1.04), and in Wilhelmi's trial, standardised mean difference = 0.08 (95% CI: −0.28–0.44). This gave an overall pooled standardised mean difference of 0.2 (95% CI: −0.07–0.44), which suggests that the control arm is unlikely to have a shorter prothrombin time.

Activated partial thromboplastin time (Fig. 6)

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References
image

Figure 6. Forest plot for activate partial thromboplastin time results. Values are activate partial thromboplastin time in seconds.

Download figure to PowerPoint

Four studies recorded this information in seconds. The normal range is 28–35 s. One study (Boldt) reported the longest times of 49.6 s for patients given FFP and 57.0 s for controls, with a standardised mean difference of −0.76 (95% CI: −1.40–0.11). Two other studies had non-significant effect sizes that favoured FFP: Kasper had a standardised mean difference of −0.23 (95% CI: −0.76–0.29), and the standardised mean difference in the Wilhelmi study was −0.29 (95% CI: −0.65–0.07). Consten found that the activated partial thromboplastin time was slightly longer in patients given FFP, with a standardised mean difference of 0.11 (95% CI: −0.44–0.67). The overall pooled standardised mean difference was just significant at −0.27 (95% CI: −0.51–0.02), with patients given FFP having a shorter activated partial thromboplastin time.

Sources of heterogeneity

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

There were no notable differences within trials in the baseline characteristics of the patients. All the studies had similar mean ages, weights and proportions of male participants. The volume of FFP infused increased in the order Boldt, Consten and Kasper, with patients in the Wilhelmi study getting at least 1000 ml of FFP if they weighed > 66.7 kg. There were no notable trends in effect with increasing dose to suggest that this affected the outcomes. The major source of bias comes from the study design, with the most bias in the Kasper, Martinowitz and Wilhelmi trials due to lack of blinding and use of FFP as a treatment during follow-up.

Discussion

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References

It is important to use FFP as efficiently as possible, not only to decrease wastage, but also to minimise the small but potential risks of acute lung injury and viral infections such as human immunodeficiency virus (HIV), hepatitis B and hepatitis C. Although there is now a very low risk of these infections (1 : 10 000 000 donations for HIV, 1 : 1 200 000 for hepatitis B and 1 : 50 000 000 for hepatitis C), there is an uncertain risk of variant Creutzfeld-Jacob disease. Tests for variant Creutzfeld-Jacob disease are currently being developed. If such a test became available, there is the risk that stocks of blood could be decreased by up to 50% if donors are reluctant to know the results [15]. Current guidelines do not recommend the prophylactic use of FFP in cardiac surgery, but do recommend its use when there is excessive bleeding associated with a documented coagulation defect that is not due to heparin [16].

We set out to identify trials in cardiac surgery that support, or do not support, this recommendation. However, no trials of therapeutic use were found, so all trials of FFP usage were reviewed. Only six randomised, controlled trials were found that aimed to evaluate its clinical efficacy when used prophylactically, and only five used a suitable control arm. The quality of the studies was fairly poor: all of the trials were small and underpowered to detect whether FFP was advantageous compared with control; the study fluid volumes were non-uniform; only one study was properly blinded; two used FFP as a treatment for all patients during follow-up, biasing the assessment of the study fluids. All studies were reported as being randomised, but only three gave a clear method of randomisation. Adequate concealment of allocation occurred in only one study, which was also the only one to perform double-blinding. However, blinding the transfusion procedure is difficult to achieve due to the yellow colour of plasma and the detailed checks that must be performed before its transfusion, which is why Consten used an independent investigator to administer the study fluid. The studies also varied in the clinical indication for administering additional FFP. In three studies (Boldt, Consten and Trimble), FFP administration was truly prophylactic and no further FFP was administered, irrespective of the degree of blood loss or the coagulation results. In one study (Wilhelmi), FFP was given only if cumulative blood loss exceeded a prespecified amount, whereas in the fourth study (Kasper), both the rate of bleeding and coagulation results had to exceed preset limits. In Martinowitz, fresh plasma was given to all control patients after the initial administration of fresh packed red cells. The very short follow-up time meant that no patients were lost to follow-up, although some were not included in the analysis of Kasper's trial because randomisation was performed before surgery, and operative complications meant that the protocol could not be followed. In terms of analyses, the poor methodological quality of the studies meant that pooling the data in a meta-analysis would not be entirely valid but, conversely, each individual trial was not large enough for the results to have an impact on clinical practice.

There was also variability in the reporting of outcome measures as described in the Methods sections. The Consten study did not give an accurate timeframe for the measurement of blood loss, nor did it report the SD of the haemoglobin concentrations. This is despite Consten being the highest quality study in terms of adequate allocation concealment and blinding, and no obvious sources of bias. Kasper only gave the median and range for blood loss, but with no measure of variation.

Possible sources of heterogeneity were addressed by looking for differences between the studies. There were few differences in the baseline characteristics of the patients, and neither the different doses of FFP nor the use of autologous plasma in two studies affected the magnitude of outcome measures. Study selection bias was possible in that all the investigators concluded that routine intra-operative administration of FFP has no benefit. This means they were more likely to be labelled ‘negative’ trials despite mainly showing small but non-significant treatment effects. However, these effects may be real and clinically important, but none of these studies had enough power to detect them.

There was an apparently beneficial effect of FFP on the platelet count in the Wilhelmi study, which was large enough to shift the standardised mean difference to near significance. However, this effect could be anomalous, as there are no platelets in FFP and no increase in platelets was seen in the other studies reviewed. Furthermore, the increased platelet count, from 130 × 109.l−1 to 158 × 109.l−1 is not clinically significant.

In conclusion, no therapeutic trials were found and none of these studies showed any benefit of administering prophylactic intra-operative FFP during coronary artery bypass surgery. The size and design, and the small numbers of subjects in these studies mean that this review is inconclusive, and will be unlikely to affect current practice until further evidence comes to light. To resolve this issue, larger trials of a higher quality are needed that specifically aim to address the circumstances under which FFP administration decreases bleeding after coronary artery surgery.

References

  1. Top of page
  2. Summary
  3. Methods
  4. Forest plots
  5. Results
  6. Trimble 1964 (California) []
  7. Boldt 1989 (Germany) []
  8. Martinowitz 1990 (Israel) []
  9. Consten 1996 (Amsterdam) []
  10. Wilhelmi 2000 (Germany) []
  11. Kasper 2001 (Germany) []
  12. Blood loss at 24 h ()
  13. Platelet count ()
  14. Fibrinogen ()
  15. Haemoglobin ()
  16. Prothrombin time ()
  17. Activated partial thromboplastin time ()
  18. Sources of heterogeneity
  19. Discussion
  20. Acknowledgements
  21. References
  • 1
    Murphy MF, Brozovic B, Murphy W, Ouwehand W, Waters AH. Guidelines for platelet transfusions. British Committee for Standards in Haematology. Transfusion Medicine 1992; 2: 3118.
  • 2
    Stainsby D, Burrowes-King V. Audits of the appropriate use of fresh frozen plasma. Blood Matters 2002; 10: 79.
  • 3
    Eagleton H, Benjamin S, Murphy M. Audits of the appropriateness of the use of fresh frozen plasma. Transfusion Medicine 2000; 10 (Suppl. 1): 6.
  • 4
    Llewelyn CA, Amin M, Ballard S et al. Transfusion medicine epidemiology pilot survey. British Journal of Haematology 2002; 117: 77 (Abstract).
  • 5
    SHOT Steering Committee. Serious Hazards of Transfusion report, July 2003.
  • 6
    Blood Transfusion Services of the United Kingdom. The Handbook of Transfusion Medicine. Norwich: The Stationery Office, 3rd edn.
  • 7
    Lewis S, Clarke M. Forest plots: trying to see the wood and the trees. British Medical Journal 2001; 322 (7300): 147980.
  • 8
    Boldt J, Kling D, von Bormann B, Zuge M, Hempelmann G. [Homologous fresh frozen plasma in heart surgery. Myth or necessity]. [German]. Anaesthesist 1989; 38: 3539.
  • 9
    Consten EC, Henny CP, Eijsman L. Dongelmans DA, van Oers MH. The routine use of fresh frozen plasma in operations with cardiopulmonary bypass is not justified. Journal of Thoracic and Cardiovascular Surgery 1996; 112: 1627.
  • 10
    Martinowitz U. Goor DA, Ramot B, Mohr R. Is transfusion of fresh plasma after cardiac operations indicated? Journal of Thoracic and Cardiovascular Surgery 1990; 100: 928.
  • 11
    Kasper SM, Giesecke T, Limpers P, Sabatowski R, Mehlhorn U, Diefenbach C. Failure of autologous fresh frozen plasma to reduce blood loss and transfusion requirements in coronary artery bypass surgery. Anaesthesiology 2001; 95: 816.
  • 12
    Wilhelmi. CABG surgery without the routine application of blood products: is it feasible? European Journal of Cardiothoracic Surgery 2001; 19: 65761.
  • 13
    Trimble AS, Osborn JJ, Kerth WJ, Gerbode F. The prophylactic use of fresh frozen plasma after extracorporeal circulation. Journal of Thoracic and Cardiovasular Surgery 1964; 48 (2): 3146.
  • 14
    British Heart Foundation. British Heart Foundation Statistics Database, 2003. 14 Fitzhardinge Street, London W1H 6DH, January 2003. http://www.bhf.org.uk.
  • 15
    Reynolds L. Donor response to a test for vCJD. Blood Matters 2001; 8: 3.
  • 16
    Contreras M, Ala FA, Greaves M et al. Guidelines for the use of fresh frozen plasma. British Committee for Standards in Haematology. Transfusion Medicine 1992; 2: 5763.