SEARCH

SEARCH BY CITATION

Keywords:

  • cardiopulmonary bypass;
  • hemorrhage;
  • thrombin generation;
  • thromboelastometry

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

Summary. Background: Bleeding after cardiopulmonary bypass (CPB) is a major cause of morbidity and mortality and consumes large amounts of blood. Identifying patients at increased risk of bleeding secondary to hemostatic impairment may improve clinical outcomes by allowing early intervention. Methods: This present study recruited 77 patients undergoing CPB and measured coagulation screens, coagulation factors, TEG®, Rotem® and thrombin generation (TG) before surgery and 30 min after heparin reversal. The tests were analyzed to investigate whether they identified patients at increased risk of excess bleeding (defined as > 1000 mL) in the first 24 h postoperatively. Results: Patients who bled > 1000 mL had a lower: platelet count (P < 0.02), factors (F)IX, X and XI (P < 0.005), endogenous thrombin potential (ETP) and an initial rate of TG (P < 0.02) and higher activated partial thromboplastin time (aPTT) (P < 0.001) than patients who bled < 1000 mL. Receiver operating characteristic (ROC) analysis was significant for post-operative TG and aPTT (P < 0.001). Furthermore, reduced pre-operative TG was associated with increased postoperative bleeding (P < 0.02). Pre- and postoperative TG were correlated (ρ = 0.7, P < 0.001). TEG®, Rotem® and prothrombin time (PT) at either time point were not associated with increased bleeding. Conclusion: These data suggest that pre-operative defects in the propagation phase of hemostasis are exacerbated during CPB, contributing to bleeding post-CPB. TG taken both pre- and postoperatively could potentially be used to identify patients at an increased risk of bleeding post-CPB.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

Bleeding after cardiopulmonary bypass (CPB) surgery is a major cause of morbidity and mortality [1,2]. Bleeding is associated with increased renal failure, cerebrovascular accident, prolonged mechanical ventilation, sepsis and a prolonged stay in critical care [3,4]. Bleeding places a major drain on blood components [5] and this may be exacerbated in the future by increasingly complex surgery in elderly patients, potentially affecting the ability of transfusion services to adequately support CPB surgery.

CPB bleeding may be caused by breeches of the vasculature, hemostatic impairment or both. Hemostatic impairment post-CPB is an interaction of dilutional and consumptive coagulopathies, increased fibrinolysis, anticoagulation and acquired defects in platelet number and function [6,7].

Current concepts of hemostasis highlight the importance of the propagation phase for generating a burst of thrombin on platelets [8]. Initiation of coagulation is through tissue factor (TF) and contact is not physiologically important although, in the context of CPB, contact contributes to depletion of coagulation factors in the extracorporeal circuits [9].

Prothrombin time (PT) and activated partial thromboplastin time (aPTT) are poor predictors of bleeding during invasive procedures [10–12] and are usually not available quickly enough to be clinically useful in a patient who is bleeding. Measuring the viscoelastic properties of coagulating whole blood by TEG® or Rotem® are useful in the management of a patient who is bleeding [1,13,14] but it is not clear whether these tests predict post-operative bleeding. Some studies have shown that abnormal TEG/Rotem predicts an increased risk of bleeding post-CPB [15–17] but this is not confirmed by others [18–20]. These inconsistencies may be explained by small numbers and different definitions of abnormal coagulation and bleeding. As a result transfusion policies are predominantly based on local custom rather than evidence [21,22].

Hemostatic assays activated by low concentrations of TF and which measure the integrated ability of coagulation factors to generate thrombin, such as calibrated automated thrombography (CAT) [23], may better reflect physiological hemostasis and hence be more useful for predicting bleeding post-CPB allowing targeted pre-emptive treatment to reduce bleeding. We assessed coagulation screens, viscoelastographic tests, CAT and individual coagulation factors taken pre- and post-CPB to investigate whether they were associated with an increased risk of bleeding during the first 24 h.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

Study design

The study was approved by the local ethics committee and informed consent was provided. Samples were taken pre-CPB and 30 mins after heparin reversal from an arterial line and the first 10 mL was discarded. The primary endpoint was the measured blood loss in the first 24 h post-surgery. Units of red bloods cells (RBC) transfused post-operatively were not used because this is influenced by intra-operative transfusion and cell salvage as well as postoperative bleeding.

Anesthetic and blood product management

Heparin was given to prolong the activated clotting time (ACT) > 480 s and reversed with protamine to achieve an ACT within 10% of baseline. Patients were cooled to between 32 and 34 °C and compound sodium lactate or Gelofusin were used as pump prime. Initially patients received aprotonin if an antifibrinolytic was required, but this agent was withdrawn during the study and subsequent patients received tranexamic acid.

Platelets were infused if < 50 × 109 L−1, fresh frozen plasma (FFP) was given for a PT/aPTT ratio > 1.5. If the fibrinogen level was < 1 g L−1 or below 1 g L−1 despite FFP, 10 units of cryoprecipitate were given. RBCs were transfused if the hemoglobin was below 6.5 g dL−1 on CPB or 8 g dL−1 postCPB [24–26]. Patients received cell-salvaged RBCs peri-operatively.

Hemostatic assays

Whole blood was tested using a TEG® 5000 Hemostasis Analyser (Haemonetics, Braintree, MA, USA) using Kaolin cups preCPB and Kaolin/Heparinase cups post-CPB and Rotem® 05 Coagulation Analyser (Tem International GmbH, Munich, Germany) using Intem and Fibtem reagents pre-CPB and Heptem and Fibtem reagents post-CPB. Fibrinogen (Clauss), thrombin time, PT and aPTT were tested on a MDAII analyser (Trinity Biotech, Bray, Ireland). Citrated platelet poor plasma (PPP), prepared by centrifugation at 3000 × g for 15 mins, was stored at −70 °C and subsequently assayed for coagulation factors (F)II, V, VII, VIII, IX, X, XI and XII on a MDAII analyser.

Calibrated automated thrombography

CAT was measured using the method by Hemker [18] on blood taken into trisodium citrate (final concentration 0.109 m) and corn trypsin inhibitor (CTI) (Cambridge BioScience, Cambridge, UK) (final concentration 20 μg mL−1) and double spun at 3000 × g for 15 mins. CAT was measured using a Fluoroskan Ascent plate reader (ThermoLabsystems, Helsinki, Finland). Fluorogenic substrate (Z-Gly-Gly-Arg-AMC) was obtained from Bachem (St Helens, UK). Phospholipid vesicles were constructed by an extrusion method and were 20% phosphatidylserine, 20% phosphatidylethanolamine and 60% phosphatidylcholine.

Eighty microliters of citrated/CTI PPP was incubated with 20 μL tissue factor solution, Innovin (Sysmex, Milton Keynes, UK) pH 7.35 diluted in HEPES/NaCl, final concentration 1 pmol. A thrombin calibrator (600 nm) was supplied by Synapse BV, Maastricht, the Netherlands. Data were analyzed using Thrombinoscope™ software (Synapse BV). The inter-assay coefficient of variances were: lag time 12%, endogenous thrombin potential (ETP) 8%, peak thrombin 8% and time to peak thrombin 10%. The rate of thrombin generation was calculated from data extracted into an Excel spreadsheet.

Statistical analysis

The primary aim of the study was to assess whether a pre- or post-operative assay could identify patients at increased risk of bleeding in the first 24 h postoperatively. The observed distribution of bleeding had a near normal distribution amongst patients who bled < 1 L and fewer patients who bleed more than 1 L (Fig. 1). This amount of blood loss has previously been suggested as a definition of excessive bleeding after CPB [6]. The assays were therefore assessed to establish whether they could discriminate between bleeding above or below 1000 mL.

image

Figure 1.  The distribution of the observed blood loss in the first 24 h post-operation is shown. Bleeds below 1000 mL were assumed to be standard and above 1000 mL defined as excess bleeding.

Download figure to PowerPoint

Data are reported as median, interquartile range (IQR) and range. Comparison of patient characteristics between those that bled more or less than 1000 mL was by either Mann–Whitney U or χ2-test. Wilcoxon’s test was used for comparisons between pre- and postsurgery and Spearman’s rank test was used for correlation with bleeding in the first 24 h postsurgery. The potential clinical utility of tests was investigated by receiver operator curve (ROC) analysis reported as area under the curve (AUC) and 95% confidence intervals (CI). In this exploratory study which aimed to find useful predictors of bleeding, no correction was made for multiple testing. Prior data were not available before the study and so a power calculation was not performed. Analyzes were performed on spss version 16 (SPSS Inc., Chicago, IL, USA).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

Seventy-seven patients were recruited, median age 69 (range 29–89) years. Operative procedures are shown in Table 1. The observed 24-h blood loss is shown in Fig. 1. This revealed a group of 55 patients with a normal distribution of blood loss who bled < 1000 mL and a group of 22 patients who bleed more than 1000 mL.

Table 1.   Characteristics and treatment of patients who bleed more or less than 1000 mL
 All patients N = 77Postoperative bleeding < 1000 mL N = 55Postoperative bleeding > 1000 mL N = 22Comparison between more or less than 1000 mL P
  1. Characteristics of patients, exposure to drugs that affect hemostasis and fluids and intra-operative blood products comparing patients who bled more or less than 1000 mL. Data are median, (IQR) and range or number and percentage. Statistical comparison is either Mann–Whitney U- or χ2-test. NS, not significant; CABG, coronary artery bypass grafts.

Bleeding during the first 24 h (mL)615 (450–1155) 130–2990525 (380–640) 130–9751665 (1341–1980) 1010–2990 
Operations
 All775522 (29%) 
 CABG24177 (29%)NS
 CABG + valve replacement332211 (33%)NS
 Valve replacement or repair862 (33%)NS
 Ascending aortic aneurysm repair541 (25%)NS
 Redo cardiac procedures761 (14%)NS
Characteristics
 Age (years)69 (57–74) 29–8967 (56–75) 29–8969 (58–73) 51–790.7
 Male/Female63/1444/1119/30.5
 Weight (kg)78 (72–88) 51–11878 (72–90) 51–11874 (66–81) 58–920.1
Management
 Aspirin exposure n (%)50/77 (65%)36/55 (65%)14/22 (64%)0.9
 Clopidogrel exposure n (%)6/77 (8%)4/55 (7%)2/22 (9%)0.8
 Antifibrinolytic usage Aprotonin n = 24 Tranexamic acid n = 4569/77 (90%)51/55 (93%)18/22 (82%)0.16
 Crystalloids (mL kg−1)12.4 (8.7–14.5) 0–38.512.4 (8.8–13.8) 0–38.512.9 (6.8–16.1) 5.4–19.00.8
 Colloid (mL kg−1)6.7 (4.3–10.9) 0–226.5 (3.2–10.9) 0–227.9 (6.4–11.0) 0–15.90.5
 Intra-operative donor red cells (units) 15 patients transfused0 (0–0) 0–30 (0–0) 0–30 (0–0) 0–20.5
 Red cell cell salvage infused (mL kg−1)8.0 (5.8–9.8) 0–15.98.0 (6.3–10.0) 0–15.96.5 (0–9.9) 0–14.40.2
 FFP (units) 6 patients transfused0 (0–0) 0–80 (0–0) 0–80 (0–0) 0–00.1
 Platelet (units) 7 patients transfused0 (0–0) 0–20 (0–0) 0–20 (0–0) 0–00.08
 Time on bypass (mins)132 (99–185) 57–324136 (101–189) 57–324130 (96–185) 81–2110.6
 Time to extubation (h)8.8 (5.9–13.8) 3–1318 (6–12) 3–13112.5 (5.7–18.4) 3.8–480.2
 Re-operative: n (%)2/77 (3%)0/55 (0%)2/22 (9%)NS
 Death n (%)110NS

Patient characteristics, peri-operative exposure to drugs that affect hemostasis and blood loss in the first 24 h postoperatively are shown in Table 1. There were no differences in exposure to anti-platelet drugs and anti-fibrinolytics or transfusion between the groups. Both groups were infused similar amounts of crystalloid, colloid and salvaged RBCs between the pre- and post-operative samples suggesting that hemostatic changes were unlikely to be predominantly as a result of dilution caused by differences in transfusion. Very few patients received FFP or platelets during the procedures and the use of these products would, therefore, have only a small effect on tests of hemostasis at the end of the procedure. Time on bypass was not associated with increased bleeding.

Association of routine coagulation tests and coagulation factor levels with post-operative bleeding

The median (IQR) and range of the PT, aPTT and fibrinogen are shown in Table 2. There was an increase in PT and aPTT and a fall in fibrinogen postsurgery compared with before. The adequate reversal of heparin in the post-operative samples is shown by the near normal thrombin clotting time in all cases.

Table 2.   Routine coagulation tests and individual coagulation factors before and after surgery and the risk of post-operative bleeding
 Pre-op N = 77 Median (IQR) RangePost-op N = 77 Median (IQR) RangePre- vs. postop PCorrelation between postop test and bleeding in first 24 h r P valuePost-op test dependent on 24 h blood lossROC analysis
Below 1 L Median (IQR) range N = 55Above 1 L Median (IQR) range N = 22PAUC (5–95% CI) P
  1. The pre-operative (pre-op) and postoperative (postop) results of routine coagulation tests and individual coagulation factors (IU dL−1) are shown. Wilcoxon’s test was used for comparison of pre- and postoperative. The ρ and P-values that assess a correlation between the post-operative sample and amount of bleeding in the subsequent 24 h are Spearman’s test. The post-operative results, dependent on whether a patient subsequently bled more or less than 1 L, are a Mann–Whitney U-test. The utility of a postoperative test to predict bleeding of more than 1000 mL is shown by receiver operator curve (ROC) and described as an area under the curve (AUC) with 95% confidence intervals (CI). NR, normal range; IQR, interquartile range; NS, not significant.

Platelet number203 (175–237) 118–654111 (85–140) 50–240< 0.001−0.33 < 0.004121 (98–144) 61–24094 (71–124) 50–1860.020.59 (0.4–0.77) NS
PT (s) NR: 11–13.513 (12.5–13.6) 10.2–15.216.1 (15–17.1) 12–20.3< 0.001NS16 (15–17) 21–2016 (15–19) 13–20NS0.42 (0.28–0.52) NS
APTT (s) NR: 26.2–36.528.7 (27.3–31.6) 23–40.543.9 (35–61) 27–109< 0.0010.43 < 0.00140 (33–54) 28–10961 (43–74) 27–89< 0.0010.75 (0.62–0.88) P = 0.002
TT (s) NR: 11–1411.3 (11–12) 10–13.512.6 (12–14) 10–25< 0.001NS12 (13–14) 10–1612 (11–14) 10–25NS0.55 (0.39–0.71) NS
Fibrinogen (g L−1)3.3 (2.7–4.1) 1.9–6.61.5 (1.2–1.9) 0.9–3.6< 0.001NS1.6 (1.3–1.9) 1.1–3.61.4 1.0–2.1 0.9–3.1NS0.57 (0.38–0.76) NS
Factor II88 (79–98) 42–13043 (35–53) 22–83< 0.001NS43.5 (36.7–53) 22–7345 (27–53) 24–83NS0.49 (0.29–0.69) NS
Factor V104 (89–115) 33–17561 (45–73) 18–132< 0.001NS64 (74–101) 24–4955 (44–72) 18–132NS0.49 (0.3–0.68) NS
Factor VII94 (79–105) 48–14855 (47–69) 26–98< 0.001NS60.5 (48.7–69) 26–9854 (41–73) 33–96NS0.62 (0.43–0.81) NS
Factor VIII124 (88–157) 32–31594 (71–127) 38–330< 0.001NS98 (124–330) 38–7180 (71–131) 44–284NS0.51 (0.33–0.69) NS
Factor IX112 (92–130) 62–18887 (68–99) 32–199< 0.001−0.33 < 0.0393 (105–156) 32–72.777 (58–88) 40–1990.0060.69 (0.52–0.86) P = 0.04
Factor X88 (77–97) 44–12941 (30–52) 13–86< 0.001−0.36 < 0.00445 (35–55) 13–8131 (23–40) 14–86< 0.0010.66 (0.49–0.83) NS
Factor XI92 (80–107) 45–14855 (42–65) 16–108< 0.001−0.40 < 0.00458 (67–96) 16–4745 (31–58) 23–1080.0050.61 (0.42–0.79) NS
Factor XII84 (68–115) 23–16845 (29–54) 8–148< 0.001NS46.5 (55–148) 8–3132 (47–110) 16–25NS0.51 (0.33–0.7) NS

There was a fall in platelets from a median of 203 × 109 L−1 pre-operatively to 111 × 109 L−1 postoperatively (P < 0.001), fibrinogen fell from a median of 3.3 g L−1 to 1.5 g L−1 (P < 0.001). The levels of all procoagulant factors fell (P < 0.001) when pre- and postsurgery samples are compared (Table 2).

Initiation phase of coagulation  When those patients that bled more or less than 1000 mL in the first 24 h postsurgery were compared, there was no difference in the FVII level.

Propagation phase of coagulation  The propagation phase of coagulation is activated by FXIa and TF/FVIIa cleavage of FIX to FIXa and driven through the tenase complex (FIXa/FVIIIa) and the prothrombinase complex (FXa/FVa). The median FXI had fallen by 60% at the end of surgery compared with pre-operatively. FXI correlated weakly with post-operative bleeding (ρ = −0.4) and the patients who bled more than 1000 mL in the first 24 h had significantly lower FXI levels than those that bled < 1000 mL. However, using ROC analysis FXI was not a useful predictive test (Table 2).

Investigation of the components of the tenase complex showed that both FIX and FVIII had fallen by the end of surgery compared with pre-operatively but almost all patients had levels of both coagulation factors that would be associated with adequate hemostasis. FIX but not FVIII was weakly correlated (ρ = −0.33) with bleeding in the first 24 h. Patients who bled more than 1000 mL had lower FIX (but not FVIII) compared with those that bled < 1000 mL. FIX level was borderline significant by ROC analysis (P = 0.04) for an association with excess blood loss.

The prothrombinase complex was also compromised with both FX and FV significantly lower postsurgery. The lowest median coagulation factor level observed was FX (41 IU dL−1). FX correlated weakly (ρ = −0.36) with bleeding in the first 24 h and was significantly lower in the patients who bled more than 1000 mL but ROC analysis was not statistically significant (Table 2).

Prothrombin, fibrinogen and platelets  The level of prothrombin was significantly reduced post-surgery with a median level of 43 IU dL−1. There was no difference in the prothrombin level when comparing those patients who would bleed more or less than 1000 mL. Fibrinogen was also reduced post surgery and the median level of 1.5 g L−1 in both groups was border line for adequate hemostasis in the context of major surgery. The levels of prothrombin and fibrinogen were the same when comparing patients who would subsequently bleed more or less than 1000 mL (Table 2).

The platelet count was reduced at the end of surgery as has been shown previously [2]. The number of platelets after surgery correlated with bleeding in the first 24 h postsurgery (ρ = −0.33) and were significantly lower in patients who would bleed more than 1000 mL during this time (P = 0.02). ROC analysis however was not statistically significant (Table 2).

Association of thrombin generation with postoperative bleeding

Postoperative sample  There was a significantly longer lag time, lower ETP and lower peak thrombin when comparing pre- and postsurgery values. There was a weak correlation between the blood loss in the first 24 h, post-operative ETP and rate of thrombin generation, although Ρ-values were low (Table 3). There was a significantly lower ETP and initial rate of thrombin generation at the end of surgery in patients who bled more than 1000 mL in the first 24 h postsurgery (Fig. 2). ROC analysis was statistically significant for ETP, peak thrombin and the initial rate of thrombin generation but the initial rate was the test with most potential clinical utility with an AUC (95% CI) of 0.76 (0.62–0.89) p = 0.004. The lag time was not a useful predictor of excess bleeding (Table 3 and Fig. 3).

Table 3.   Thrombin generation before and after surgery and the risk of postoperative bleeding
 Pre-op median (IQR) mangePostop median (IQR) mangePre- vs. postop P-valueCorrelation pre-op test and bleeding in first 24 h ρ PCorrelation postop test and bleeding in first 24 h ρ PPre-op test dependent on 24 h blood lossPost-op test dependent on 24 h blood lossROC analysis
Below 1000 mL Median (IQR) RangeAbove 1000 mL Median (IQR) RangePBelow 1000 mL Median (IQR) RangeAbove 1000 mL Median (IQR) RangePPre-op tests AUC (95% CI)Post-op tests AUC (95% CI)
  1. The pre-operative (pre-op) and postoperative (post-op) results of thrombin generation tests are shown. The P-values that compare the pre- and postoperative samples are calculated using Wilcoxon’s test. The ρ and P-values that assess a correlation between the pre- and postoperative sample and amount of bleeding in the subsequent 24 h are calculated using Spearman’s rank test. The pre- and postoperative results, analyzed dependent on whether a patient subsequently bled more or less than 1000 mL, are compared using a Mann–Whitney U-test. The utility of the pre- and postoperative tests to predict bleeding of more than 1000 mL is shown by receiver operator curve (ROC) and described as an area under the curve (AUC) with 95% confidence intervals (CI). NS, not significant.

Lag time (min)3.8 (3.2–4.2) 2.3–7.75.3 (6.9–4.3) 2.6–13< 0.001NS  NS3.9 (3.3–4.4) 2.4–7.33.6 (3.1–4.1) 2.5–7.8NS5.3 (4.4–7.1) 3–13.685.2 (3.6–7.0) 2.7–8.9NS0.56 (0.41–0.71) NS0.56 (0.39–0.73) NS
ETP (nmol min−1)1059 (859–1427) 351–2321981 (611–1296) 20–2154< 0.001−0.26 0.03−0.28 0.021091 (905–1526) 378–2321949 (752–1123) 351–15490.031025 (652–1386) 42–2154684 (344–1003) 20–15070.020.68 (0.54–0.82) P = 0.030.71 (0.57–0.85) P = 0.009
Peak (nmol)145 (97–190) 17–345135 (74–201) 1.8–2850.03−0.25 0.03  NS155 (100–213) 48–345134 (75–158) 17–2060.03155 (88–217) 1.8–28689 (31–146) 2.3–241NS0.67 (0.54–0.81) P = 0.030.69 (0.54–0.83) P = 0.02
Initial rate (nmol min−1)35 (21–46) 3–10729 (12–55) 0.38–116  NS−0.26 0.03−0.3 0.0138 (22–53) 11–10730 (16–39) 3.3–530.0345 (18–68) 0.8–11617 (4.0–32) 0.4–550.010.67 (0.54–0.81) P = 0.030.76 (0.62–0.89) P = 0.004
image

Figure 2.  The median, interquartile range and range of endogenous thrombin potential (ETP) and initial rate of thrombin generation (nmol min−1) measured at the end of surgery in patients who would subsequently bleed more or less than 1000 mL is shown. Comparison is made with a Mann–Whitney U-test, *P < 0.02.

Download figure to PowerPoint

image

Figure 3.  Receiver operator curve (ROC) for utility in predicting bleeding of more than 1000 mL by thrombin generation parameters post-operation.

Download figure to PowerPoint

Pre-operative sample  Although pre-operative thrombin generation was within the population normal range there was an association between the pre-operative ETP, peak thrombin, rate of thrombin generation and blood loss in the first 24 h post-CPB (Table 3 and Fig. 4). This finding was confirmed using ROC analysis with all three parameters having utility for identifying patients at increased risk of excess post-CPB blood loss (Table 3).

image

Figure 4.  Median, interquartile range and range of endogenous thrombin potential (ETP) (nmol min−1) and initial rate of thrombin generation measured pre-operation in patients who would subsequently bleed more or less than 1000 mL (Mann–Whitney U-test *P < 0.03).

Download figure to PowerPoint

The pre- and postoperative ETP correlated strongly (ρ = 0.7, P < 0.001) suggesting that those with low normal thrombin generation pre-operatively were more likely to have critically reduced thrombin generation postsurgery and this that contributed to the increased risk of bleeding (Fig. 5).

image

Figure 5.  Endogenous thrombin potential (ETP) before and after cardiopulmonary bypass (CPB) (Spearman’s test, ρ = 0.7 P < 0.001).

Download figure to PowerPoint

Association of viscoelastic properties of clot with postoperative bleeding

Thromboelastometry assays were performed on the first 44 patients recruited. There was a significant change in all parameters of both TEG and Rotem at the end of surgery compared with presurgery. However, none of the parameters correlated with the amount of bleeding in the first 24 h postsurgery. Comparison of the 26 patients who bled < 1000 mL with the 18 that bled more than 1000 mL in the first 24 h postsurgery demonstrated no difference between any parameters. None of the parameters demonstrated a statistically significant association with blood loss in the first 24 h on ROC analysis (Table 4). Because no utility for predicting blood loss was observed at the interim analysis after 44 patients, these tests were not performed on the subsequent cohort.

Table 4.   Viscoelastic properties of clots before and after surgery and the risk of postoperative bleeding
 Pre-op median (IQR) mange N = 44Postop median (IQR) range N = 44Pre vs. postop PCorrelation postop test and bleeding in first 24 h ρ PPostop test dependent on 24 h blood lossROC analysis
Below 1 L median (IQR) range N = 26Above 1 L median (IQR) range N = 18P valueAUC (95% CI) P
  1. The pre-operative (pre-op) and postoperative (post-op) results of viscoelastic clotting tests are shown. The P values that compare the pre- and postoperative samples are calculated using Wilcoxon’s test. The r and P values that assess a correlation between the postoperative sample and amount of bleeding in the subsequent 24 h are calculated by a Spearman test. The postoperative results, dependent on whether a patient subsequently bled more or less than 1000 mL, are compared using the Mann–Whitney U-test. The utility of a postoperative test to predict bleeding of more than 1 L is shown by receiver operator curve (ROC) and described as an area under the curve (AUC) with 95% confidence intervals (CI). NS, not significant.

TEG: Kaolin TEG cups for pre-op and kaolin heparinase cups for postop
 R time (min)7.3 (5.5–8.4) 2.2–158.3 (6.1–12.6) 2.8–40< 0.001NS7.6 (6.5–14.4) 4.2–406.8 (5.7–11.7) 2.8–40NS0.55 (0.37–0.73) NS
 K time (min)1.9 (1.3–2.3) 0.8–4.62.2 (1.8–2.9) 0–12.5< 0.001NS2.2 (1.8–3.1) 0–12.52.2 (1.5–2.8) 0–10.9NS0.58 (0.41–0.76) NS
 Alpha angle (degrees)65 (57–70) 28–8160 (49–65) 0–76< 0.03NS58 (42–63) 0–7061 (51–70) 0–76NS0.43 (0.23–0.61) NS
 MA (mm)70 (67–75) 60–8460 (53–65) 0–73< 0.001NS60 (54–64) 0–7260 (51–65) 0–73NS0.51 (0.33–0.69) NS
Rotem: Intem for pre-op and heptem for post-op
 Clot time (s)156 (139–178) 127–280204 (175–238) 124–2400< 0.001NS211 (190–242) 156–635189 (167–241) 124–2400NS0.63 (0.45–0.81) NS
 Clot formation time (s)64 (57–77) 40–102142 (109–199) 0–433< 0.001NS138 (111–204) 71–322157 (91–200) 0–433NS0.51 (0.32–0.70) NS
 Maximum clot firmness (mm)63 (59–65) 54–7551 (44–54) 0–64< 0.001NS52 (45–56) 40–6247 (52–54) 0–64NS0.59 (0.42–0.78) NS
 Alpha angle (degrees)77 (75–78) 70–8264 (56–69) 0–79< 0.001NS65 (57–69) 40–7663 (54–70) 0–79NS0.53 (0.34–0.72) NS
Fibtem for both pre and post op samples
 Maximum clot firmness (mm)17 (12–19) 8–449 (7–11) 4–27< 0.001NS9 (7–11) 4–279 (6–13) 4–24NS0.49 (0.29–0.69) NS

Association of routine coagulation tests with postoperative bleeding

There was a significant difference between the pre- and postsurgery PT and aPTT (Table 2). There was no difference in the PT when comparing the patients who bled more or less than 1000 mL. There was, however, a statistically significant difference in the end of surgery aPTT in patients who bled more or less than 1000 mL in the subsequent 24 h. The potential clinical utility of the aPTT was confirmed on ROC analysis with an AUC (95% CI) of 0.75 (0.62–0.88) P = 0.002.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

These data demonstrate that patients who bled more than 1000 mL in the first 24 h post-CPB had fewer platelets and lower FXI, FIX and FX at the end of surgery compared with those who bled < 1000 mL. The integrated effect of all procoagulant factors, measured by thrombin generation [23], was better that individual coagulation factor levels for identifying patients at increased risk of excess bleeding. Remarkably, pre-operative thrombin generation also identified patients at increased risk of excess post-operative bleeding.

These findings suggest that a defect in the propagation phase of hemostasis is implicated in excess bleeding post-CPB and this gives a new insight into the pathophysiology of bleeding in this clinical situation. Lower FXI would reduce activation of FIX and this, associated with the lower FIX and FX levels, would be expected to combine to reduced FXa generation and explain the reduced thrombin generation in the patients with excess bleeding. Thrombin generation was found to be more useful for identifying patients at increased risk of bleeding than individual coagulation factors probably because it measures the integrated effect of the changes in these factors on coagulation [23]. Thrombin generation tests in platelet-rich plasma have not been reported because many patients were thrombocytopenic and it was not possible to standardize the platelet count in the assay.

Fibrinogen was similar in the groups that bled more or less than 1000 mL, about 1.5 g L−1; however, there is increasing evidence that this is a level that could be considered borderline for major surgery [27]. In the present study, decreased thrombin generation may not have been able to drive the formation of a sufficiently stable fibrin clot in patients with borderline fibrinogen levels.

In contrast, changes in assays that measure the viscoelastic properties of clots were not associated with subsequent post-operative bleeding in the present study. This finding both reproduces and contradicts other studies [16–19]. A possible reason for the observed differences is that fibrinogen levels are important in clot-based but not thrombin generation assays and, because fibrinogen was similar in the patients that bled more or less than 1000 mL, this may have overwhelmed the ability of these assays to detect defects in the propagation phase of coagulation. The low concentration of TF and contact inhibition in CAT makes the assay much more sensitive to the propagation phase. The findings support the view that thrombin generation and viscoelastic clot-based assays are sensitive to different aspects of hemostasis.

The clinical utility of thrombin generation assays in the context of CBP surgery is yet to be defined. At present these assays are used in specialized laboratories and delays in obtaining results limits their potential clinical use. However, if the development of near patient whole blood thrombin generation assays is successful measurement would be potentially useful in identifying patients at increased risk of bleeding and who might benefit from more proactive replacement of coagulation factors. This will require further studies designed to identify possible thrombin generation values that can be used to trigger intervention and assessment of the effect of an intervention. In addition, thrombin generation assays may be useful in providing early information to assess whether patients who are bleeding post CBP have a hemostatic defect or a potential reversible surgical cause of bleeding.

It is not known whether correction of the defect in the propagation phase of hemostasis and hence thrombin generation, for example by infusion of FFP or prothrombin complex concentrates or over-riding it with activated recombinant factor VII (rFVIIa), would have prevented the excess bleeding and this will require prospective randomized studies. A randomized study with rFVIIa in patients bleeding post-CPB reduced bleeding but increased critical adverse events [28]. A test such as thrombin generation, that identifies patients at increased risk of bleeding as a result of hemostatic impairment, will be useful in future studies by allowing therapy to be targeted at the patients who most need it and reduce the risk of thrombotic complications by tailoring the replacement dose to the hemostatic defect [28,29].

The finding that thrombin generation measured pre-operatively identified patients at increased risk of post-operative bleeding was unexpected. No previous pre-operative hemostatic assay has been shown reproducibly to be predictive of post-operative bleeding [12]. A possible explanation is that patients with thrombin generation in the lower part of the normal range were more susceptible to developing coagulopathies associated with CPB compared with patients with higher pre-operative thrombin generation who would have more hemostatic reserve. This suggestion is supported by the finding that pre- and postoperative thrombin generation is strongly correlated. These data suggest that thrombin generation is a more sensitive test of hemostasis than coagulation screens, at least in the context of CPB, and that people with thrombin generation at the lower end of the normal range may be more prone to bleeding after a major hemostatic challenge.

The primary endpoint of observed drain loss was chosen because it can be directly measured. An alternative measurement such as post-operative red blood cell transfusion is not directly related to post-operative bleeding because transfusion is based on hemoglobin and is therefore heavily influenced by intra-operative bleeding and transfusion of cell salvaged and donor red cells. Increased post-operative transfusion is also a late indicator of bleeding and less sensitive than directly observing drain loss. Intra-operative bleeding is less well related to hemostasis than postoperative bleeding because the high levels of heparin virtually inactivated hemostatic mechanisms. Although hemostatic assays were shown to predict a proportion of post CPB bleeding in the present study they are not sensitive to surgical bleeding and cannot, therefore, be expected to be predictive in all cases.

Hemostatic defects associated with CPB have been shown previously to be related to consumption of coagulation factors and platelets in the circuit, dilution, activation of fibrinolysis and a transient platelet function defect [6,7]. Previous studies have reported that the length of time on CBP is a risk factor for bleeding [3] but this was not observed in the present study, indeed decreased thrombin generation in the normal range was associated with increased post-operative bleeding before bypass had occurred.

The proportion of patients experiencing a 1000-mL blood loss was higher before the interim analysis than after. It is unclear why this was found and it is most likely as a result of biological variation. The only apparent change in treatment in the latter phase compared with the initial phase of the study was the change from aprotinin to tranexamic acid; however, this is an unlikely reason for bleed rates to fall.

As thrombin generation assays have become more available [23] numerous studies have been published [30]. There have, however, been very few studies that have linked thrombin generation to bleeding endpoints. The present study is proof that thrombin generation has a potentially useful role to detect clinically important hemostasis impairment more sensitively than other assays.

In conclusion, these data demonstrate that thrombin generation measured both before and after CPB can identify patients at increased risk of post-operative bleeding. The important hemostatic defect appeared to be in the propagation phase of hemostasis. Whether these assays will be clinically useful or correction of this defect will reduce bleeding and the associated morbidity will require prospective study.

Addendum

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

M. Coakley: study design, patient recruitment, sample collection, performance of near patient testing, data interpretation, manuscript writing. J. E. Hall, A. R. Wilkes, P. W. Collins: study design, data interpretation and manuscript writing. C. Evans, E. Duff, V. Billing, L. Yang, D. McPherson: patient recruitment, sample collection, performance of near patient testing, data interpretation, manuscript writing. E. Stephens, N. Macartney: laboratory testing, data interpretation and manuscript writing.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

The study was support, in part, by the British Heart Foundation and The Welsh Blood Service. The Rotem machine and reagents were provided by Pentapharm.

Disclosure of Conflict of Interests

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References

The authors state that they have no conflict of interest.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References