The overall prevalence of acute seizures in patients > 60 yr of age is estimated to be 0.1%.1,2 Traditionally, the causes of seizures in elderly patients include acute stroke (50%), metabolic encephalopathy (6–30%), and various drug side effects, infection, and trauma (5–20%).3,4 The incidence of seizures after cardiac surgery is not well studied.5,6 There have been recent reports, however, of postoperative seizures in adult patients who had been given tranexamic acid (TA) as antifibrinolytic therapy.7-10 Although the reported seizures were generally easily treated, the implications of a seizure diagnosis in this age group are significant, including restrictions on normal activities (e.g., driving) that may affect the patient’s quality of life, independence, and self-confidence.11

We recently observed what appeared to be an increase in postoperative seizures after cardiac surgery. Prior to January 2007, our antifibrinolytic strategy involved ε-aminocaproic acid (and rarely TA) for patients at low risk for bleeding (e.g., uncomplicated first-time coronary artery bypass surgery) and aprotinin for high-risk patients (e.g., redo cardiac surgery, aortic surgery, multivalve operation). In 2007, the Blood Conservation Using Anti-fibrinolytics in a Randomized Trial (BART) showed an increased morbidity and mortality associated with the use of aprotinin.12 Partly as a result of this trial, the use of aprotinin was removed from clinical practice. Many institutions, including our own, changed their use of antifibrinolytic agents to TA alone. Tranexamic acid can cross the blood-brain barrier (BBB), especially in circumstances where the BBB has been compromised. This property had previously been exploited in neurosurgery to decrease the risk of rebleeding following subarachnoid hemorrhage. However, its use in neurosurgery was subsequently discontinued after TA was shown to be associated with increased cerebral vasospasm, ischemia, hydrocephalus, and seizures.13 Tranexamic acid binds competitively, in a dose-dependent fashion, to γ-aminobutyric acid (GABA) type A receptors, which results in reduced inhibitory activity and increased neuronal excitation.14 There are reports of TA causing seizures when in direct contact with the central nervous system in animals15,16 and humans.17,18 It is therefore mechanistically plausible that TA is associated with seizures after cardiac surgery.

We sought to determine the risk factors for seizures after cardiac surgery and to characterize them. We specifically examined TA as a potentially modifiable factor.

Methods

We conducted a historical cohort study using data collected in our cardiac surgery, perfusion service, and intensive care unit (ICU) databases. These databases collect information on all consecutive cardiac surgery cases performed at our institution. Each database undergoes random audits to ensure validity. The perfusion service database collects detailed operative variables (including the antifibrinolytic agent that was used during surgery). The cardiac surgery and ICU databases collect demographic and co-morbidity variables along with postoperative complications, duration of ventilation, and lengths of ICU and hospital stays. A concurrent pharmacy database maintains information related to the inventory of drugs dispensed to various clinical services, including anesthesia for cardiac surgery. The University of Manitoba Research Ethics Board approved the study and waived the requirement for patient consent.

A seizure was deemed to have occurred if the attending physician documented the occurrence and a prescribed intervention occurred. Our databases were searched for the occurrence of postcardiac surgery seizures from April 2003 (when the perfusion service database was available) until December 2009. A further chart review of the patients with seizures was then undertaken to identify factors associated with seizures (e.g., preoperative stroke, preoperative medications), the characteristics of the seizure, and the in-hospital treatment and outcome. Patients with seizures were compared to those without seizures to determine the factors associated with seizures.

The potential contribution of TA was assessed in two ways. First, TA was examined as a dichotomous variable in a multivariable logistic regression analysis to determine its association with seizures. Second, a dose–response relation for TA was examined. The TA dose was calculated per case based on pharmacy dispensing records of the amount of TA dispensed divided by the number of patients receiving TA. Prior to 2007, our institutional policy was to administer 1–2 g of TA per patient (for those prescribed TA). With the BART results,12 our institutional policy changed in 2007 to a 30 mg·kg−1 bolus followed by 16 mg·kg−1·hr−1 for the duration of cardiopulmonary bypass (CPB). Although we did not have the exact dose per patient available for all patients (owing to inconsistencies in recording dosing in the anesthesia record), when we cross-referenced these institutional standards with our pharmacy dispensing records (amount of TA dispensed divided by the total cases performed in which TA was prescribed), the doses were remarkably similar, giving us confidence in the dosing data.

Statistical analysis was performed using SPSS 17.0 (SPSS, Chicago, IL, USA). The preoperative characteristics of the two groups (seizure, non-seizure) were compared using Student’s two-sample t test (for normally distributed data), the Mann-Whitney U-test (for non-normally distributed data), and Pearson’s Chi-squared test (for categorical variables). Multivariable logistic regression, using a forward stepwise approach, was performed on all clinically relevant and statistically significant variables available from the initial bivariate analysis (Table 1) to determine factors associated with seizures after cardiac surgery. The variables examined in the regression analysis included the following: TA use; age > 70 yr; history of preoperative neurological disease; preoperative cardiogenic shock; preoperative cardiac arrest; Acute Physiology, Age, and Chronic Health Evaluation (APACHE) score > 20; peripheral vascular disease; preoperative renal insufficiency; preoperative aortic valve disease; preoperative mitral valve disease; preoperative aortic disease; previous cardiac surgery; CPB time > 150 min; open chamber procedure. All reported P values are two-tailed.

Table 1 Bivariate analysis of potential predictors of seizure after cardiac surgery: April 2003 to December 2009
Full size table

Results

There were 56 patients with seizures documented among the 5,958 cases (0.94%). The baseline rate of 0.19% for the period April 2003 to December 2006 went up 8.1-fold to 1.54% for the period January 2007 to December 2009 (P < 0.001). Table 1 demonstrates that patients who experienced seizures were older and had more preexisting neurological disease, arrhythmias, cardiac arrests, congestive heart failure, previous cardiac surgery, and renal dysfunction, in addition to having higher APACHE II scores. They also had more open-chamber procedures, longer cross-clamp and CPB times, and were more frequently prescribed TA. Table 2 shows the unadjusted odds ratio (OR) and the final multivariable model with the adjusted ORs demonstrating that TA use, preoperative cardiac arrest, APACHE II score > 20, preoperative neurological disease, open-chamber procedure, CPB time > 150 min, and previous cardiac surgery were significant factors for postoperative seizure (Nagelkerke pseudo R2 = 0.27, P < 0.001).

Table 2 Logistic regression analysis of cardiac surgery patients having seizures: April 2003 to December 2009
Full size table

The Figure demonstrates the changes in TA dosing over time and the change in the seizure rate over time. From April 2003 to December 2006, patients who were given TA received 1.4–1.7 g per case, and the seizure rate during that period was 0–0.3%. From January 2007 to December 2009, patients given TA received 5.1–5.8 g per case, and the seizure rate was 1.2–1.8%. Our initial analysis at the end of 2009 showed that patients who had seizures received 80.9 ± 24.2 mg·kg−1 (mean ± SD) of TA vs 63.9 ± 24.1 mg·kg−1 of TA (P < 0.001) for the patients who did not have seizures. Therefore, an institutional policy change was implemented wherein the total amount of TA prescribed was decreased to < 45 mg·kg−1 per case. This number was chosen based on a pharmacokinetic modeling study done by Dowd et al.19 that suggested adequate antifibrinolytic activity with a 12.5 mg·kg−1 bolus, an infusion of 6.5 mg·kg−1·hr−1 and 1 mg·kg−1 in the pump prime (i.e., 26.5 mg·kg−1 for an average 80-kg patient on CPB for 2 hr), and the BART protocol12 (which would give 64 mg·kg−1 to an 80-kg patient on CPB for 2 hr). By averaging the two protocols, we arrived at the administered TA dose of 45 mg·kg−1. The Figure shows a reduction in the TA dosing to an average of 3.1 g per case (from January to September 2010) and a simultaneous reduction in the seizure rate from 1.8% to 0.8%. Importantly, from January 2007 to December 2009 compared to January 2010 to September 2010, the APACHE II score (14.0 ± 4.8 vs 14.1 ± 4.3), preoperative cardiac arrest rate (0.5% vs 0.7%), preoperative neurological disease rate (13% vs 14%), and previous cardiac surgery rate (1.8% vs 2.3%) did not change. Furthermore, the number of open-chamber procedures increased (from 32.5% to 37.5%) as did the CPB time (108.8 ± 58.6 vs 113.0 ± 63.6 min); the increase in open-chamber procedures was clinically and statistically significant (P = 0.01). This suggests that the reduction in seizures was likely related to a reduction in the TA dose, not a change in the other significant factors from the multivariable model.

Figure
figure 1

Change in seizure incidence and in tranexamic acid (TA) dosing over time. The TA dose was calculated per case based on the pharmacy records and number of patients given TA. N = the total number of cardiac surgery cases done during that time period. The graph shows the increase in the incidence of seizures after cardiac surgery starting in 2007 correlated with the increase in TA dosing. Note the subsequent drop in the seizure rate in 2010 when the TA dose was decreased

Full size image

To investigate the impact of TA on seizures further, we performed two additional analyses. First, we examined the data for all of the patients who were given TA and all who were not and determined the seizure rate differences between the two groups. The seizure rate in patients not receiving TA was 0.2% (similar to the baseline rate of 0.1% in patients > 60 yr),1,2 and the seizure rate in patients receiving TA was almost sevenfold higher, at 1.47% (P < 0.001). Second, we examined the data for patients with none of the other six risk factors from our model for seizures (patients with no previous neurological history, no prior cardiac arrest, no previous cardiac surgery, and no open-chamber procedure but with an APACHE score < 20 and a CPB time ≤ 150 min). This produced a sample size of 2,786 patients, 1,596 of whom had been given TA (three seizures occurred in this group) and 1,190 who had not (zero seizures occurred in this group). Our data suggest that TA may influence the probability of seizures after cardiac surgery independent of the other risk factors.

Table 3 shows the characteristics of the 61 patients who had seizures (56 before December 31, 2009 and five thereafter). Most seizures occurred within 6 hr of surgery and were grand mal in nature; 58% recurred during the hospital stay. However, almost half of the patients had their anticonvulsant medications discontinued within 3 days. The incidence of postoperative neurological complications, defined as delirium and/or stroke, were higher in the seizure group (19.6% vs 3.2%, P < 0.001), as were ICU length of stay [4.7 (2.0–7.8) vs 1.0 (0.8–2.0) days, P < 0.001] and ICU mortality (9.7% vs 1.4%, P = 0.001). Interestingly, the three patients who had seizures and only had TA as a seizure risk factor (i.e., none of the other six risk factors) did not have neurological complications, a prolonged ICU length of stay, or any mortality.

Table 3 Characteristics of patients with seizures after cardiac surgery (56 patients before TA dose change; 5 patients after TA dose decrease)
Full size table

Discussion

In this large single-center study, we found an association between the occurrence of seizures after cardiac surgery and the following factors: TA use, preoperative cardiac arrest, high APACHE II score, preoperative neurological disease, open-chamber procedure, long CPB time, and previous cardiac surgery. We also demonstrated a dose–response relationship that suggests an increase in seizures with an increase in TA dose and a decrease in seizures after a decrease in TA dose per case. Although the absolute difference in TA dosing between the seizure and non-seizure groups was not large, it is possible that the circulating and brain levels of TA were much higher in the seizure group because that group was older and had renal dysfunction (reducing TA elimination).

It is noteworthy that most of the seizure events were within hours of arriving in the ICU, and most were grand mal in nature. This timing likely represents the relatively high plasma levels of TA (which we did not measure) present at that time, resulting in reduced GABA-A receptor activity. This, combined with the weaning of sedative drugs to allow early extubation, structural brain abnormalities in some patients (as demonstrated by computed tomography of the head), and possible BBB abnormalities due to intracerebral embolic events from the open-chamber procedure, may have allowed the proconvulsant properties of TA to be unmasked. Even though many patients had recurrent seizures, almost half of the patients were able to discontinue anticonvulsant medication use prior to discharge. This further supports the notion that transient factors (e.g., TA) play an important role in precipitating seizure activity in susceptible patients. It appears that TA may precipitate seizures in patients who have other significant risk factors for seizure.

Clinically, the problem of TA-associated seizures can be addressed in several ways. First, TA could be withdrawn from use during cardiac surgery until it is subjected to large clinical trials versus placebo to assess its effect on all-cause mortality and morbidity. The BART12 found increased mortality and morbidity with aprotinin versus TA, leading to the removal of aprotinin from clinical practice. However, the question remains whether TA, in adequately powered clinical studies versus placebo, improves all-cause outcomes. The second option would be to continue using TA until such a large outcome trial is performed but to minimize the risk of TA-induced seizures. Although reducing the TA dose appears prudent, what constitutes an effective TA dose is unclear. It is possible that TA has a relatively flat dose–response curve or that the dosing of TA is clinically relevant only in patients at risk for seizure based on the other factors found in our model. Some limited pharmacokinetic data suggest that a minimum therapeutic TA plasma concentration of 127 μM is required.19 Dowd et al.19 used a TA regimen of 12.5 mg·kg−1 given over 30 min, a maintenance infusion of 6.5 mg·kg−1·hr−1, and 1 mg·kg−1 in the CPB prime, which maintained a TA concentration > 345 μM in most patients. The BART protocol12,19 was predicted to give a TA concentration of > 800 μM using this same pharmacokinetic modeling study. Fox20 reported a dosing regimen of a 5.4 mg·kg−1 load followed by 5 mg·kg−1·hr−1 with 20 mg·L−1 in the CPB pump prime, with no reported seizures and adequate antifibrinolytic activity as measured by thromboelastography. Santos et al.21 performed a double-blinded randomized control trial in cardiac surgery patients with very low dose TA (10 mg·kg−1 load followed by 1 mg·kg−1·hr−1 for 5 hr) versus no antifibrinolytic agent and found that D-dimer levels were significantly higher in the control group than in the TA group, but there was only a modest decrease in blood loss (approximately 300 mL less blood per case in the TA group at 24 hr postoperatively).

With the uncertainties that exist regarding the effective TA dose, modifying the perioperative course could be considered in some at-risk patients (as described by the regression model). This could include further minimizing the TA dose, being meticulous about cardiac de-airing, minimizing aortic manipulation, and potentially continuing sedative medications during the postoperative period until TA blood levels are predicted to be low. Similarly, in situations where a higher dose of TA is deemed required (e.g., redo sternotomy, thoracic aortic surgery, complex cardiac surgical procedures, surgery requiring hypothermic circulatory arrest), higher doses of GABA agonists could be considered during and immediately after surgery.

Our findings support those currently in the literature and add important new information. In three studies7,22,23 that primarily investigated the hemostatic effects of aprotinin versus TA, patients receiving aprotinin had fewer seizures (0–1.2%) than those having received TA (2.7–4.6%).7,22,23 It is noteworthy that seizures were not reported by the BART investigators, although seizure was not an a priori defined endpoint in that trial.12 There have been two recent reports that have specifically examined TA and seizures after cardiac surgery. Murkin et al. described seizures in 669 cardiac surgery cases in London, Ontario and Cambridge, UK.10 Their baseline seizure rate of 1.3% increased to 3.8% when higher TA dosing was adopted. The TA dose varied in each of the centres: 61–259 mg·kg−1 and 71–258 mg·kg−1, respectively. Similar to our experience, when the TA dose was reduced (in their case, to that of the BART protocol),12 the seizure rate decreased. As in our series, most of the patients were elderly, seizures were predominantly grand mal in nature, they occurred within hours of surgery, and most of the patients had open-chamber procedures. There was documented evidence of stroke in 12% of their patients; but unlike our series, no recurrent seizures were reported. In another series, Bell et al. reported seven patients with seizures after cardiac surgery.9 Their baseline seizure rate of 0% increased to 0.66% with high-dose TA (60 mg·kg−1 bolus followed by 5 mg·kg−1·hr−1 or a 30 mg·kg−1 bolus followed by 15 mg·kg−1·hr−1). The patients were elderly, all patients underwent open-chamber procedures, and the seizures were grand mal in nature. All of the seizures occurred within 24 hr, and there were no recurrences after 24 hr. The reason for the high recurrence rate of seizures in our population is unclear at this point but warrants further investigation.

There were some limitations to our study. There are the inherent problems related to database and chart reviews. We did have some missing data (5–10% of random variables); however, this degree of missing data is comparable to that of other retrospective studies and would be unlikely to affect the conclusions of our study. We reviewed the charts of the patients who had a seizure in detail and did not find any biochemical abnormality or medications (e.g., lidocaine, bupivacaine, cephalosporins, others) or evidence of central nervous system infection or trauma that could readily explain the seizures. All patients were given standard preoperative antibiotics (usually 1–2 g of cefazolin), which has been the routine for years; thus, it is likely that this was not the reason for the increased seizures. For logistical reasons, we could not do a detailed review of the charts of the several thousand patients who did not have seizures to determine their biochemical and medication profiles. With all the patients being managed by a homogeneous single-center group of clinicians, and the fact that the seizure rate dropped after decreasing the TA dose, it seems unlikely that other biochemical derangements or medications explained the initial increase and subsequent decrease in the seizure rate.

Ideally, one would have documented seizure activity by electroencephalography (EEG) in all these patients to confirm that the movements seen were indeed seizures and not nonseizure movement. Most seizures lasted only a few minutes, thus precluding the ability to perform an urgent confirmatory EEG during the seizure activity. However, the nursing and medical staff working in the ICU, all experienced practitioners, were accustomed to seeing the common nonseizure involuntary movements after surgery (e.g., shivering) and were therefore likely accurate in their clinical diagnoses. Additionally, the same nursing and medical staff were working in the ICU after the TA dosing was reduced (without them being informed of the dosing change), and the subsequent incidence of seizures was reduced.

Tranexamic acid was used less frequently before 2007 but was the sole agent used from 2007 onwards. Our analysis examined two time periods (2003–2006 and 2007–2009). Thus, it is possible that our study may have biases associated with before/after studies (e.g., selection, history, information bias). However, we do not think the probability of this is high as other groups have reported results similar to ours,10 with higher TA doses leading to increased seizures. It was not possible to examine the role of TA from 2007 to 2009 as all patients received TA thereby eliminating a no-TA group for comparison. Our dose–response relation to seizure rates attempts to compensate for this limitation.

In conclusion, we have addressed risk factors associated with seizures after cardiac surgery. Notably, TA may be a significant and readily modifiable factor in patients with other risk factors for seizure. We suggest that because of the association of TA with seizures one should use TA in the lowest possible antifibrinolytic dose, particularly in patients with other risk factors for seizure. For patients deemed to be at high risk for seizures, as determined by our multivariate model, but in whom high-dose TA may be necessary, consideration should be given to infusing GABA agonists (propofol or midazolam) during the initial high-risk period.