Postoperative Bleeding in Cardiac Surgery: The Role of Tranexamic Acid in Patients Homozygous for the 5G Polymorphism of the Plasminogen Activator Inhibitor-1 Gene

This was a secondary analysis of a randomized, double-blind, placebo-controlled study to investigate the effect on blood loss of a single dose of TA (2 g) administered before and after elective CPB surgery.9The main study was discontinued by the hospital ethics committee (University Hospital of the Canary Islands, Tenerife, Spain) before completion of enrollment due to a higher rate of excessive bleeding in the placebo group. The study population consisted of 50 patients classified according to PAI-1 genotype (4G/4G, 4G/5G, or 5G/5G). Twenty-four received 2 g TA before and after CPB, whereas 26 received placebo. In this report, we investigate the secondary endpoint of the influence of patient PAI-1 genotype on TA acid efficacy.

Inclusion criteria were consecutive white adult patients undergoing elective CPB surgery. Exclusion criteria included chronic coagulopathy (prothrombin time <50% or international normalized ratio >2, and platelets <50,000/mm³or aggregation dysfunction), renal dysfunction (creatinine >2 mg/dl), hepatopathy (Child B or higher degree), or use of immunosuppressant drugs. Before CPB, participants had normal bleeding time, platelet collagen/epinephrine and collagen/adenosine diphosphate closure time, prothrombin time, activated partial thromboplastin time, and thrombin time. The patients did not receive antiinflammatory agents or anticoagulant therapy (acetyl salicylate acid, heparin, warfarin, antiplatelet therapy) for 5 days before surgery and for the first 24 h after surgery. The study was approved by the hospital ethics committee, and all patients gave written informed consent for genetic analysis and enrollment in the clinical trial.

Anesthetic procedures were standardized and consisted of an opioid-based anesthetic supplemented with volatile agents and muscle relaxants. The extracorporeal circuit consisted of a hard-shell membrane oxygenator (Optima XP; Cobe, Denver, CO, or Quantum Lifestream International, Inc., Woodlands, TX), a Tygon® extracorporeal circuit (Saint-Gobain, Charny, France), and a Medtronic Biopump® centrifugal pump (Tolochenaz, Switzerland). The patients received heparin to achieve an activated clotting time of greater than 480 s. Body temperature was decreased to 28°–30°C during CPB, and the pump flow was adjusted to maintain a mean arterial pressure of greater than 60 mmHg and a flow index of 2.2 l · min⁻¹· m⁻². St. Thomas 4:1 blood cardioplegia was used at 12°C. Heparin was reversed with protamine with a dose based on blood heparin levels measured by Hepcon® HMS (Medtronic, Minneapolis, MN). A blood salvage device was used in all patients. Based on our clinical practice, the patients received packed erythrocytes when hemoglobin was less than 8 g/dl. When postoperative blood loss was deemed clinically excessive, the patients received fresh frozen plasma when the prothrombin time was less than 50% (reference range, 70–100%) or the international normalized ratio was greater than 2, and a platelet transfusion when the platelet count was less than 50,000/mm³.

Demographic information, comorbid conditions, and other clinical data were recorded. The patient surgical risk was calculated by Parsonnet and Acute Physiology and Chronic Health Evaluation II scores. Chest tube blood loss was measured at admission to intensive care unit (0 h) and at 4 and 24 h after surgery.

Quantitative PAI-1 antigen (range, 2–47 ng/ml; intraassay variation, 3.7%) and tissue plasminogen activator antigen (tPA) (range, <9.0 ng/ml; intraassay variation, 4.2%) were measured using enzyme-linked immunosorbent assay (Imubind® America Diagnostica Inc., Stanford, CT). An automated chromogenic method (Berichrom® Dade Behring Inc., Newark, NJ) was used for the quantitative determination of the functional activity of antithrombin (reference range, 80–130%).

Blood was collected before surgery into EDTA-containing glass tubes. DNA purification was conducted using proteinase K, phenol-chloroform extraction, and ethanol precipitation. The DNA samples were stored at 4°C in Tris-EDTA buffer, and genotyping was performed in a double-blinded manner. Analysis of PAI-1 gene polymorphism was performed using primers and restriction endonuclease digestion as previously described.10This polymorphism is due to the insertion of an extra G in the 5G allele, which also generates a restriction site. Thus, while the 4G allele is only 99 bp long, the 5G allele is indeed 100 bp long in the initial amplification, but it becomes 77 + 23 as a result of restriction enzyme treatment. Briefly, primers PAI-F: CACAGAGAGAGTCTGGCCACGT and PAI-R: CCAACAGAGGACTCTTGGTCT were used to amplify a 99-bp fragment, which was subject to digestion with restriction enzyme BslI. The 5G PAI-1 allele, which contains the BslI recognition sequence (CCN7GG), yields 77- and 23-bp fragments, whereas the 4G allele (resistant to BslI) remains as a 99-bp fragment. Restriction fragments were resolved by 5% acrylamide gel electrophoresis. Because spurious genetic associations can result from the existence of genetic substructure in the population sampled, we also genotyped 22 neutral genetic markers (polymorphic sites in the genome that are considered to be functionally neutral) as a genomic control strategy. If the associations were due to population substructure, some of the unrelated, neutral markers would be expected to yield significant association with the phenotype studied also. The neutral markers chosen were Alu repeats and short tandem repeats distributed throughout the genome.

Quantitative variables are reported as median and interquartile range. Nominal variables are reported as frequency and percentage. Assumption of normality of the TA and placebo groups was tested with the Kolmogorov–Smirnov test, and homoscedasticity was tested with the Levene test. First, comparisons between placebo and TA groups were conducted. The Pearson chi-square or Fisher exact test was used to compare sex, type of surgery, erythrocyte and plasma requirements, and Parsonnet score using the Kruskal–Wallis test for single-order categories. Comparison between these groups on age, platelets, hemoglobin, international normalized ratio, PAI-1 levels, CPB and clamping time, temperature, total heparin and protamine dose, blood salvage device, and chest tube blood loss was performed with the Student t  test or Mann–Whitney U test as appropriate. Second, patients were divided into three groups of PAI-1 genotypes (4G/4G, 4G/5G, and 5G/5G), except when comparing TA effect on 24-h bleeding, when two groups were compared (5G/5G vs.  other PAI-1 genotypes). Patients classified according to their genotype were compared on redo using the Kruskal–Wallis test for single-order categories. The Fisher exact test was used for comparing groups on sex, type of surgery, and perioperative parameters (CPB and clamping time, total heparin, protamine dose, blood salvage device, hematocrit, prothrombin time, international normalized ratio, fibrinogen, antithrombin, and tPA). PAI-1 genotypes, PAI-1 levels, and blood loss are quantitative variables (ordinal and continuous respectively). A previous study has reported that there is a natural ordering of PAI-1 levels according to PAI-1 genotypes.11We applied the Jonckheere–Terpstra test to determine whether there was a linear relation in a double natural ordering. It is used to compare “r” populations (PAI-1 genotypes), defined by different levels of a quantitative variable (PAI-1 protein levels and chest tube blood loss), where each level contains subjects whose quantitative responses can be classified into “c” distinct response levels.

An exact Kruskal–Wallis test was used to compare transfusion requirement differences according to PAI-1 polymorphism, and an exact Mann–Whitney U test was used to study the effects of TA on bleeding endpoints for each genotype. Given the small sample size, to adjust for potential confounding variables, the effect of the PAI-1 polymorphism on bleeding was analyzed after pooling the 4G carriers together. Twenty-four–hour chest tube blood loss was log transformed to normalize the distribution in order to perform an analysis of covariance test of 5G/5G versus  4G carriers and controlling for redo, combined surgical procedure (known risk factors for bleeding),12and body mass index (significant demographic variable in the univariate analysis), in three independent models. To assess the possible existence of population substructure, we conducted a logistic regression analysis using as independent variables the neutral markers and as the dependent variable the treatment group, as suggested by Setakis et al.  13Only one marker was included in each logistic regression analysis. Genotypes were tested for Hardy–Weinberg equilibrium by using Online Encyclopedia for Genetic Epidemiology software.∥Hypotheses were two-tailed. Statistical significance was defined as a P  value less than 0.05 for comparisons between placebo and TA groups. An uncorrected P  value was established as 0.05. The Bonferroni correction with a corrected P  value was applied for a posteriori  comparisons between genotypes and independent variables as follows: demographic variables, P < 0.0125; hematocrit, P < 0.025; PAI-1 and tPA levels, P < 0.0125; coagulation parameters such as prothrombin time, fibrinogen, and antithrombin, P < 0.0125; chest tube blood loss, P < 0.017; and transfusion requirements during intensive care unit stay (erythrocyte, plasma, and platelets units), P < 0.017. We used SPSS version 12 (SPSS Inc., Chicago, IL) and StatXact 5.0.3 (Cytel Software, Cambridge, MA).

Of the 70 patients initially enrolled, only 50 were randomly assigned to receive either TA (n = 24) or placebo (n = 26). Exclusion reasons were “off-pump” surgery (n = 5), coagulopathy (n = 2), emergency surgery (n = 5), endocarditis (n = 4), Jehovah's Witness faith (n = 1), hemodialysis (n = 1), or protocol violation (n = 2).

Baseline demographic and other patient data were similar between TA and placebo groups. Bivariate analysis showed that the TA group had significantly lower chest tube blood loss at all postoperative time points as compared with the placebo group. In addition, the TA group showed significantly lower transfusion requirements of erythrocytes and plasma during their intensive care unit stay (table 1).

There were no differences between the TA and placebo groups with respect to neutral markers (data not shown). The 4G/4G genotype was found in 10 patients (20%), the 4G/5G was found in 26 (52%), and the 5G/5G was found in 14 (28%) of the 50 patients studied. The distribution of PAI-1 polymorphism was similar to that observed in 100 consecutive blood donors from our institution, in Hardy–Weinberg equilibrium.

The perioperative data grouped according to PAI-1 polymorphism are shown in table 2. There were differences in serum levels of PAI-1 between the three PAI-1 genotypes, before surgery (4G/G: 49 [33–56] ng/ml, 4G/5G: 34 [25–48] ng/ml, and 5G/G: 28 [17–35] ng/ml; P = 0.046; corrected P  nonsignificant) and to intensive care unit admission (4G/G: 67 [46–97] ng/ml, 4G/5G: 61 [36–82] ng/ml, and 5G/5G: 34 [29–42] ng/ml; P = 0.018; corrected P  nonsignificant). Significant differences were found between PAI-1 genotype groups (4G/4G, 4G/5G, and 5G/5G) in chest tube blood loss at 0 h (30 [28–45], 114 [88–215], 160 [98–250] ml, respectively; P = 0.03; corrected P  nonsignificant), at 4 h (160 [98–210], 253 [185–574], 661 [505–1,045] ml, respectively; P = 0.001; corrected P  significant), and at 24 h (428 [313–548], 715 [475–1,175], 1,220 [1,045–1,855] ml, respectively; P = 0.009; corrected P  significant) after surgery.

The 5G/5G genotype was associated with 24-h chest tube blood loss, controlling for reoperation (F^1,26= 9.48, P = 0.005), combined surgical procedures (F^1,26= 11.29, P = 0.003), and body mass index (F^1,26= 6.36, P = 0.02).

Fresh frozen plasma was required during the intensive care unit stay in 50% of 5G/5G, 25% of 4G/5G, and none of 4G/4G carriers (P = 0.021; corrected P  nonsignificant). There were no differences in other transfusion requirements (erythrocyte units and platelets).

Tranexamic acid decreased chest tube blood loss in 5G carriers. 4G/4G patients did not show significant differences in blood loss between TA and placebo groups. 4G/5G patients receiving TA had lower blood loss than placebo group at 0 h (60 [40–93]vs.  114 [88–215] ml, respectively; P = 0.012; corrected P  significant) and at 24 h after surgery (325 [213–770]vs.  715 [475–1,175] ml, respectively; P = 0.014; corrected P  significant). In contrast, 5G/5G patients receiving TA had significantly lower blood loss compared with the placebo group at 0 h (55 [25–70]vs.  160 [98–250] ml, respectively; P = 0.028; corrected P  nonsignificant), at 4 h (140 [100–152]vs.  661 [505–1,045] ml, respectively; P = 0.008; corrected P  significant), and at 24 h (293 [160–357]vs.  1,220 [1,045–1,855] ml, respectively; P = 0.004; corrected P  significant) after surgery (fig. 2). Fifty-five percent of 5G/G patients in the placebo group received fresh frozen plasma during their intensive care unit stay compared with no patients in the TA group (P = 0.014; corrected P  significant). There were no differences in platelet requirements.

Our results show that the efficacy of TA prophylaxis for reducing postoperative chest tube blood loss was dependent on patient PAI-I genotype. 5G carriers treated with TA had significantly lower blood loss than those receiving placebo; however, homozygous 5G patients showed the greatest descent in blood loss. In contrast, the efficacy of TA prophylaxis to reduce blood loss compared with placebo was not observed in homozygous 4G patients.

Bleeding after CPB surgery is multifactorial4,5; however, there is growing evidence that genetic factors influence postoperative bleeding. The fact that certain patients have disorders in hemostasis that only manifest in response to major insults such as cardiac surgical procedures14,15supports this view.

Many studies demonstrate that the 4G allele in the 4G/5G polymorphism of the PAI-1 gene is associated with higher plasma PAI-1 levels, whereas patients with 5G alleles have low PAI-1 levels.7In the current study, PAI-1 levels were measured by a validated method.16,17We found differences between the three PAI-1 genotypes in the levels of PAI-1, although these differences were not significant. Recently, a correlation has been demonstrated between PAI-1 genotype and messenger RNA transcription in CPB surgery.184G homozygotes had the highest plasma PAI-1 levels, heterozygote subjects were intermediate, and 5G homozygotes had the lowest levels of PAI-1. It seems that PAI-1 polymorphism is a determinant factor with respect to serum PAI-1 levels. Thus 5G homozygotes may be predisposed to greater bleeding due to insufficient inhibition of fibrinolysis. It remains possible, however, that the association of the 4G/5G polymorphism with the observed phenotype reflects only the role of an unknown gene nearby, in linkage disequilibrium with PAI-1.

Antifibrinolytic drugs have been used as prophylaxis for postoperative bleeding,19but there is no universally accepted consensus on what strategy to use and in which patients.20A relevant finding of the current study was the identification of patients with greater tendency to bleeding (5G carriers, particularly homozygotes) who benefit most from antifibrinolytic prophylaxis.

This study has limitations. Because of the sample size, multivariable models could not be adequately conducted, and this renders several confounding scenarios as potential explanations. Body mass index has been associated with postoperative bleeding15,21as well as with the PAI-1 polymorphism in several reports.22,23We could not properly study the interaction between body mass index and PAI-1 polymorphism on perioperative parameters; nevertheless, homozygous 5G was associated with increased 24-h postoperative bleeding after adjusting for body mass index. Although sample size is the main limitation of the study, we assessed the validity of the magnitude of the effects of the differences in chest tube blood loss between genotypes using neutral markers and the Hardy–Weinberg equilibrium test. We found a significant association between genotype (5G/5G) and clinical repercussion (postoperative bleeding and transfusion requirements). Further studies with a greater number of patients are required to confirm these findings.

5G carriers showed greater tendency to post-CPB chest tube blood loss. 5G homozygotes, not receiving TA, showed significantly more postoperative bleeding than patients with other PAI-1 genotypes. 5G homozygotes who received TA showed the greatest blood-sparing benefit.

The authors thank their hospital laboratory colleagues, especially Juan Sanchez, M.D. (Staff, Molecular Biology Department, University of La Laguna, Tenerife, Spain), Eduardo Salido, M.D., Ph.D. (Professor, Molecular Biology Department, University of La Laguna), and Alejandro Jiménez, M.D. (Statistician, Research Unit, University Hospital of the Canary Islands, Tenerife, Spain).