Administration of Tranexamic Acid Improves Long-Term Outcomes in Total Knee Arthroplasty

Abstract

Background:

Blood transfusion in total knee arthroplasty (TKA) is associated with increased morbidity, including periprosthetic joint infection (PJI). Tranexamic acid (TXA) reduces blood transfusion rates, but there is limited evidence demonstrating improved outcomes in TKA resulting from TXA administration. The objectives of this study were determining whether TXA was associated with decreased rate of PJI, decreased rate of outcomes associated with PJI, and whether there were differences in rates of adverse events.

Methods:

A multicenter cohort study comprising 23,421 TKA compared 4,423 patients receiving TXA to 18,998 patients not receiving TXA. Primary outcome was PJI within two years of TKA. Secondary outcomes included revision surgery, irrigation and debridement, transfusion, and length of stay. Adverse events included readmission, deep vein thrombosis, pulmonary emboli, myocardial infarction, or stroke. Adjusted odds ratios were determined using linear mixed models controlling for age, sex, thromboembolic prophylaxis, Charlson comorbidity index, year of TKA, and surgeon.

Results:

TXA administration reduced incidence of PJI by approximately 50% (OR 0.55; p=0.03). Additionally, there was decreased incidence of revision surgery at two years (OR 0.66; p=0.02). Patients receiving TXA had reductions in transfusion rate (OR 0.15; p<0.0001) and length of stay (p<0.0001). There was no difference in the rate of pulmonary emboli (OR 1.20, p=0.39), myocardial infarction (OR 0.78, p=0.55), or stroke (OR 1.17, p=0.77).

Conclusion:

Administration of TXA in TKA resulted in reduced rate of PJI and overall revision surgery. No difference in thromboembolic events were observed. The use of TXA is safe and improves outcomes in TKA.

Level of Evidence, Level III, Observational Cohort Study

Introduction

Allogenic blood transfusions in knee and hip arthroplasty are associated with an increased risk of severe adverse events. Morbidity and mortality associated with these transfusions include increased length of stay[1], deep venous thrombosis (DVT)[2], wound healing problems[3], and ninety-day mortality [4]. Further, the overall long term survival of the implant is at risk as allogenic transfusions have been associated with increased rates of periprosthetic joint infection (PJI) [5].

These concerns have led to a focused effort to decrease the rate of perioperative transfusions in knee and hip arthroplasty. The expanded use of anti-fibrinolytics, primarily tranexamic acid (TXA), has resulted in a large decrease in allogenic transfusions in arthroplasty procedures. Over the last decade, transfusions have decreased from about one third of all knee replacements in 2010 [6] to a current rate of approximately 2% [7–18]. This overwhelming evidence at the efficacy of TXA has led to its widespread application into practice. The ultimate goal of TXA is to reduce the complications associated with transfusions such as PJI. Further, there were initial concerns that as an antifibrinolytic agent, administration of TXA may increase the risk of postoperative vascular occlusive and thromboembolic events. As the treatment effect of TXA is large, these clinical studies have needed smaller patient numbers to demonstrate the primary end point of a reduced transfusion rate. This has made it difficult to observe differences in smaller more rare events such as periprosthetic joint infection or other adverse vascular occlusive complications such as pulmonary emboli, myocardial infarction, or stroke.

The objective of this study was to determine if the use of TXA improves long term outcomes or changes the risk of adverse events in total knee arthroplasty. As TXA reduces transfusion rates, we hypothesized that there would be a corresponding decrease in transfusion-related complications. To overcome previous limitations in observing these rare events, we designed a large multicenter retrospective observational study to quantify differences in the rates of PJI, revision surgery, and vascular occlusive events.

Methods

Study Design

A multicenter observational retrospective study was completed of all patients undergoing a primary total knee arthroplasty between 2005 and 2015. Institutional review board approval was obtained, and data were acquired from the electronic medical records and manually validated for accuracy. The study was completed at 16 hospitals in a regional health system compromising a variety of academic, private practice, and hospital-employed environments in rural and urban settings.

We performed an a priori calculation to determine the sample size required to detect a statistically significant difference in the rates of our primary outcome, PJI, and main adverse events, myocardial infarction and stroke. For PJI, we assumed a baseline rate of 1% for the group of patients not receiving TXA [19]. For myocardial infarction and stroke, we assumed a baseline rate of 0.3% for the group of patients not receiving TXA [20]. Furthermore, we assumed a 4:1 ratio of sample sizes between groups given the initial distribution in adoption of TXA at these institutions. In order to observe an approximate 50% reduction in the rate of PJI to 0.5 % in the TXA group and conclude a significant between-group difference, we would need 16,000 patients that did not receive TXA and 4,000 patients that did receive TXA. In order to observe an approximate doubling in the incidence of myocardial infarction or stroke to 0.7% in the TXA group and conclude a significant between-group difference, we would need a similar distribution of patients as calculated in our PJI sample size calculations. These calculations assumed a two-tailed z-test for the difference between two independent proportions, as well as 90% power and a type I error rate of 0.05.

Participants

A patient flow diagram is provided demonstrating identifying participants, exclusion criteria, and distribution between treatment groups (figure 1). The initial cohort was obtained querying the medical record for all total knee arthroplasties completed at the 16 participating institutions identifying 26,705 patients. Laterality of the procedure was obtained for each primary total knee arthroplasty. Exclusion criteria included no data on intraoperative or perioperative medications (3,272), PJI diagnosis on same date as TKA (n=3), PJI diagnosis preceding the date of the TKA (n=6), Revision TKA prior to the date of the primary TKA (n=2), and abortion of TKA procedure for intraoperative medical condition (n=1).

Figure 1.

Flow of patients through the study

Description of Treatment

A total of 23,421 TKA were then analyzed to compare primary and secondary outcomes and adverse events. In this cohort, 4,423 patients were administered tranexamic acid during the TKA procedure and 18,998 patients were not administered TXA. As multiple studies have demonstrated no difference in transfusion rates between topical, intravenous, or oral administration of TXA, these were pooled into one group [22–26]. Of those undergoing TKA who were administered TXA, 61% were intravenous, 23% were oral, and 16% were topical.

Outcome Measures, Data Sources, and Bias

Irrigation and debridement procedures were defined as any deep or superficial procedure following the TKA that was defined by the surgeon as a debridement. Revision procedures were defined as any subsequent surgical procedure after the index TKA.

Diagnosis of TKA PJI was determined by the surgeon and identified using the electronic medical record. Periprosthetic joint infection was defined using the International Classification of Disease-9 code (ICD-9 code) for periprosthetic joint infection (996.66). As this diagnosis code can be used for different anatomic arthroplasty implants, TKA PJI was identified as the intersection of all patients in the medical system with the diagnosis code and a concurrent surgical procedure on the knee, as determined by the medical record.

Adverse events

Diagnosis of thromboembolic and vascular occlusive events were made by the treating surgical team and supportive consultant services, and identified using the electronic medical record.

The adverse events of pulmonary emboli, myocardial infarction, deep vein thrombosis, and stroke were identified with their ICD-9 codes using electronic medical record data (Pulmonary emboli: 415.19, 415.1; Myocardial infarction: 410; Deep vein thrombosis: 453.40, 453.41; Stroke: 434, 436). Each of these events were then confirmed by direct review of the medical record to verify the event occurred within the appropriate time frame and met diagnostic criteria.

Statistical Analysis

Descriptive statistics were computed to quantify demographic and clinical characteristics of patients in this study. Means and standard deviations were used for continuous variables with symmetric distributions, and medians and interquartile ranges were reported for continuous variables with skewed distributions. Categorical variables were summarized using counts and percentages.

Comorbid conditions of patients were identified using ICD-9 codes obtained from electronic medical billing records. The Charlson Comorbidity Index (CCMI) was calculated for each patient using the conventional method [27]. Weighted counts of a patient’s comorbid conditions were used to compute the widely-used combined comorbidity index.

Chi-square and t-tests were employed to compare these characteristics between those who received TXA and those who did not, for univariate analysis. Multivariate analysis was completed by using generalized linear mixed models with a logit link and random surgeon effect to test for differences in the proportion of TKAs that had an adverse event between those who received TXA and those who did not while adjusting for age, gender, CCMI, aspirin (ASA), and whether or not the surgery was performed before 2010. This time point was selected as it was the midpoint in the study. For the continuous outcome LOS, a generalized linear mixed model with a log link was used with the same random and fixed effects. All management of electronic medical record data and statistical analyses were performed using SAS Version 9.4.

Results

Demographics and Description of Study Participants

Demographics of the study participants were described (Table 1). Of the 23,421 cases of TKA that met inclusion criteria during the study period, 4,423 received TXA and 18,998 did not (Figure 1). There were no differences in gender or the incidence of diabetes and rheumatoid arthritis. The cohort that received tranexamic acid were older, received relatively less aspirin, and had a lower CCMI score. As adoption of the use of TXA in our system began in 2011, there was an expected difference in the use of TXA before and after 2010.

Table 1.

Patient demographic and characteristics.

	Total (n=23,421)	TXA (n=4,423)	No TXA (n=18,998)	p-value
Age (years), mean (SD)	66.5 (9.8)	67.5 (9.2)	66.3 (10.0)	<0.0001
CCMI, median (IQR)*	0 (2.0)	0 (1.0)	0 (2.0)	<0.0001
Female, N (%)	14,704 (62.7)	2,787 (63.0)	11,917 (62.7)	0.73
Diabetes, N (%)	1,479 (6.3)	273 (6.2)	1,206 (6.4)	0.67
Rheumatoid Arthritis, N (%)	535 (2.3)	89 (2.0)	446 (2.4)	0.18
TKA performed before 2010, N (%)	4,333 (18.5)	4 (0.1)	4,329 (22.8)	<0.0001
ASA, N (%)	4,555 (19.5)	729 (16.5)	3,826 (20.1)	<0.0001

^*Charlson comorbidity index calculated excluding age as a factor.

Abbreviations: SD, standard deviation; CCMI, Charlson comorbidity index; IQR, interquartile range; ASA, aspirin

Primary and Secondary Outcomes

The use of tranexamic acid improved long-term outcomes following total knee arthroplasty (Table 2). The adjusted outcomes controlled for several factors including the use of aspirin, CCMI, age, gender, surgeon, and time of procedure. For the primary outcome, the use of tranexamic acid reduced the incidence of periprosthetic joint infection by approximately 50% in the first two years following surgery (OR 0.55, [95% CI: 0.42–0.94]; p<0.05). For secondary long-term outcomes, the rate of revision surgery at 2 years was decreased by more than 30% in those patients receiving TXA compared to those who did not (OR 0.66, [CI: 0.46–0.93]; p<0.01). There was no difference in the rate of irrigation and debridement at 2 years post-TKA between the groups (OR 1.20, [CI: 0.62–2.32]; p=0.59).

Table 2.

Primary and secondary outcomes.

	TXA (n=4,423)	No TXA (n=18,998)	Unadjusted Odds Ratio (95% CI)	p-value	Adjusted Odds Ratio (95% CI)	p-value
PJI 2 yrs from surgery	29 (0.7)	239 (1.3)	0.51 (0.20–0.84)	0.014	0.55 (0.42–0.94)	0.033
I&D 2 yrs from surgery	15 (0.3)	116 (0.6)	0.97 (0.51–1.86)	0.93	1.20 (0.62–2.32)	0.59
Revision rate at 2 yrs	46 (1.0)	432 (2.3)	0.59 (0.42–0.83)	0.0027	0.66 (0.46–0.93)	0.019
Transfusion at 1 week	63 (1.4)	2,195 (11.6)	0.15 (0.11–0.19)	<0.0001	0.15 (0.11–0.19)	<0.0001
LOS, median (IQR)	3.0 (0.6)	3.2 (1.2)	0.91* (0.89–0.93)	<0.0001	0.93* (0.91–0.95)	<0.0001

^*Note: This value is not an odds ratio, but a rate ratio. Abbreviations: TXA, tranexamic acid; CI, confidence interval; PJI, periprosthetic joint infection; I&D, irrigation and debridement; yrs, years; LOS, length of stay; IQR, interquartile range. Statistically significant values are highlighted.

The use of tranexamic acid improved short-term outcomes following total knee arthroplasty (Table 2). The use of tranexamic acid reduced transfusion rates from 12 to 1% (OR 0.15, [CI: 0.11–0.19]; p<0.0001). There was a 7% decrease in length of stay in those patients receiving TXA as compared to those who did not (CI: 0.91–0.95; p<0.0001).

Adverse Events

There were no differences in adverse events observed between administering or not administering tranexamic acid intraoperatively in total knee arthroplasty (Table 3). There was no difference in the rate of thromboembolic or vascular occlusive events between these two groups. This included deep vein thromboses (OR 0.92, [CI: 0.45–1.89]; p=0.82), pulmonary emboli (OR 1.2, [CI: 0.80–1.79]; p=0.39), myocardial infarction (OR 0.78, [CI: 0.34–1.77]; p=0.55), and stroke (OR 1.17, [CI: 0.41–3.37]; p=0.77). There was no difference in readmission rates between these two groups either at 30 days (OR 1.02, [CI: 0.86–1.20]; p=0.83) or 90 days (OR 1.03, [CI: 0.89–1.20]; p=0.68).

Table 3.

Adverse events.

	TXA (n=4,423)	No TXA (n=18,998)	Unadjusted Odds Ratio (95% CI)	p-value	Adjusted Odds Ratio (95% CI)	p-value
DVT at 30 days	11 (0.2)	47 (0.2)	0.87 (0.43–1.76)	0.70	0.92 (0.45–1.88)	0.82
PE at 30 days	39 (0.9)	119 (0.6)	1.37 (0.92–2.04)	0.12	1.20 (0.80–1.79)	0.39
MI at 30 days	8 (0.2)	50 (0.3)	0.70 (0.32–1.53)	0.37	0.78 (0.34–1.77)	0.55
Stroke at 30 days	5 (0.1)	17 (0.1)	1.12 (0.40–3.16)	0.82	1.17 (0.41–3.37)	0.77
Readmission 30 days	246 (5.6)	1,384 (7.3)	0.94 (0.80–1.10)	0.44	1.02 (0.86–1.20)	0.83
Readmission 90 days	327 (7.4)	1,748 (9.2)	0.95 (0.82–1.09)	0.44	1.03 (0.89–1.20)	0.68

Abbreviations: TXA, tranexamic acid; CI, confidence interval; DVT, deep vein thrombosis; PE, pulmonary embolus; MI, myocardial infarction. There were no statistically significant values.

Discussion

Allogenic blood transfusions in total knee arthroplasty are associated with a small increased risk of adverse long-term outcomes, including periprosthetic joint infection [28–30]. Multiple orthopaedic studies have demonstrated that tranexamic acid reduces blood transfusion rates [31–35], but these studies have not demonstrated a reduction in the risks that are associated with these transfusions. Further, as the efficacy of tranexamic acid at reducing transfusion rates is large, these well-designed clinical studies have been smaller as compared to traditional drug-based studies investigating the safety profile of a pharmaceutical agent. We completed a large multicenter retrospective observational study to quantify the long-term benefits of TXA and determine if there was an increased incidence of these more difficult to observe adverse events. Tranexamic acid was associated with a reduction in periprosthetic joint infection and its associated revision surgery, and there was no observed increased risk in thromboembolic or vascular occlusive events.

The use of tranexamic acid in total knee arthroplasty improves long-term outcomes, as demonstrated by reduced rates of periprosthetic joint infection and revision surgery. In patients that did not receive tranexamic acid, the rate of early periprosthetic joint infection was 1.3%. This is a comparable value to other large studies that have demonstrated that infection remains the primary cause of early failure in total knee arthroplasty [36–42]. In our cohort, the use of tranexamic acid reduced the rate of observed PJI by 50%. Other groups have observed a trend in the reduction in PJI with the use of TXA by a comparable magnitude, but given the smaller population size, this reduction was not statistically significant. We were further able to observe a decrease in overall 2-year revision surgical rates by more than 30%. As PJI is the most common reason for early failure and revision surgery in TKA, it is not surprising that a reduction in PJI observed by using TXA was also associated with a large reduction in revision surgery.

The use of tranexamic acid was observed to improve perioperative outcomes in TKA. It has been well documented that tranexamic acid reduces the rate of transfusion in TKA. Initially multiple retrospective studies [31–33, 43–45] and then prospective studies documented a large reduction in transfusion rates [34, 46, 47]. The range of reported transfusion was approximately 20 to 30% without the use of TXA. Addition of TXA reduced transfusion rates to approximately 1% [48]. Our study observed similar results and serves as an important control demonstrating the accuracy of our data for longer-term outcomes. We also observed a small but statistically significant reduction in length of stay by 7%. This is likely the result that acute post-operative anemia is typically diagnosed towards the end of the in-patient stay and requires additional monitoring during and after the transfusion. Late diagnosis of anemia would likely delay discharge.

Tranexamic acid is an antifibrinolytic agent, and its mechanism of action in reducing blood loss is based on preventing the disruption of thrombus formation. There was an initial careful introduction of TXA based on safety concerns as thrombus formation is associated with pulmonary emboli, myocardial infarctions, and stroke. Numerous retrospective and prospective orthopaedic clinical studies have not observed an increase in thromboembolic and vascular occlusive events with TXA [49]. These high-quality studies were designed to detect differences in transfusion rates with the use of TXA. These studies had decreased patient numbers as the sample size needed to demonstrate efficacy was small as compared to the large size needed to observe an adverse event. Our a priori sample size calculation suggested that approximately 20,000 TKA would be needed to observe an approximate doubling in the incidence of myocardial infarction or stroke, assuming a reported rate of approximately 0.3% [20]. We did not observe a significant difference in thromboembolic events with or without the use of tranexamic acid with a sample size of 23, 421.

However, there are several limitations in this study that are important to consider. As with similar observational studies, there are inherent limitation including the retrospective nature of data collection and lack of randomization between treatment cohorts. We attempted to control for this by completing a multivariate analysis of patient-specific factors that are known to contribute to the risk of PJI. When this was completed, TXA continued to be an independent predictor of reduced incidence of PJI and revision rates. Additionally, as this dataset was collected from a large administrative database, we were unable to control for incomplete documentation or coding errors which other large database studies encounter. To account for this, each primary outcome, secondary outcome, and adverse event was manually confirmed within the medical record for accuracy. Unfortunately, body mass index was not consistently recorded within the medical record and hence was not included in our analysis. Furthermore, we were unable to control for surgeon specific perioperative antibiotic regimen, including use of topical vancomycin powder, dilute betadine lavage, or use of antibiotic impregnated cement. It is reasonable to consider that these factors could have influenced the incidence of PJI. Other confounders that should be considered include differences in perioperative wound care, route of TXA administration, post-operative rehabilitation protocol, and discharge disposition. Lastly, 80% of the non-TXA cohort underwent TKA while TXA was available, which may represent lack of familiarity with current best practices in arthroplasty. This confounder is likely due to the variations in hospital settings that patient data was extracted from. Thus, a multitude of factors contribute to the incidence of adverse events after TKA and it cannot be definitively concluded that TXA alone is responsible for reduction in PJI and revision surgery. Nonetheless, we believe that our results suggest that TXA is a major contributing factor to reductions in these postoperative complications.

Tranexamic acid has been widely adopted in total knee arthroplasty given its ability to reduce transfusion rates with no observed increase in adverse complications. Transfusions are associated with decreased long-term outcomes including PJI. We have demonstrated that TXA reduces the rate of PJI by almost 50% and its associated revision surgery. We did not observe a significant difference in adverse thromboembolic or vascular occlusive events. This provides important evidence for its continued widespread use. Continued evaluation, monitoring, and prospective studies are needed.