Patients older than 18 years were eligible if they were treated with a NOAC (rivaroxaban, apixaban, edoxaban, or dabigatran), were scheduled for a dental extraction, and provided written informed consent. Patients who were pregnant or lactating were excluded [11]. Patients visiting the oral and maxillofacial departments of the following Belgian hospitals were screened for eligibility: University Hospitals Leuven, Regional Hospital Heilig Hart Leuven, General Hospital St-Jan Genk, and AZ Monica Antwerp.

All patients were instructed to skip the morning dose of their NOAC on the day of the dental extraction, in agreement with the European Hearth Rhythm Association (EHRA) guidelines [6] and as previously validated by our group [1]. No other precoagulation screening was performed, in agreement with the EHRA guidelines [6]. If the procedure was planned early in the day, it was also allowed to skip the anticoagulant dose the evening before, aiming for an 18 to 24 hours window before the last dose. Patients were randomized through an interactive web recognition system to a 1g/10 mL (10%) TXA mouthwash or a taste and color-matching placebo, both manufactured by the Leuven Centre for Clinical Pharmacology, University Hospitals Leuven, Belgium. A computer-generated block randomization list was generated by an independent person (Leuven Clinical Coordinating Centre, University of Leuven, Belgium) for treatment allocation. The randomization was not stratified. Patients were enrolled in the study and assigned to the study interventions by clinicians. Patients, clinicians, and treating surgeons were blinded to the allocated intervention.

Patients were instructed to use the mouthwash immediately before the dental extraction and subsequently 3 times a day for 3 days starting the day after the extraction. They were asked to use the mouthwash for 1 minute and to spit it out afterwards. Extraction wound management was left at the discretion of the physician. Patients were instructed to resume the NOAC the day after dental extraction unless there were hemostasis issues. Investigators blinded to the allocated treatment contacted the patients by phone on day 2 and day 7 after dental extraction to assess compliance with the study protocol, the occurrence of any bleeding, and other secondary and safety outcomes [11].

The primary outcome was the number of patients with any oral bleeding, defined as overt bleeding within the oral cavity. Secondary outcomes were the number of bleeding events in the different bleeding categories, periprocedural bleeding (defined as a bleeding score measured on a visual analogue scale ranging from 0 to 10), unplanned medical contacts, reinterventions following oral bleeding, and unplanned NOAC interruptions. Oral bleeding events were categorized as minor, clinically relevant, or major, as previously published [1], and as early (on the day of the extraction or the day after) or delayed (on day 2 or later) [11]. Facial hematomas were recorded as well. A blinded adjudicator adjudicated all suspected bleeding events and reinterventions. Safety outcomes were thrombotic events, including myocardial infarction, stroke, systemic embolism, and venous thromboembolism, any allergic reactions to the mouthwash, and non-oral bleeding events (such as epistaxis, eye bleeding, or bruises), up to the end of the study follow-up [11].

The study was designed to test the hypothesis that TXA mouthwash would be superior to placebo with respect to the number of patients with any oral bleeding. Based on a pilot study, the sample size was set at 236 patients, which would allow to detect a treatment difference of 15% between the TXA and placebo group using a chi-squared test with an expected proportion of patients with any bleeding of 30% in the control group [1]. The statistical power was set at 80% and the 2-sided significance level at 5%. All analyses were performed using SAS version 9.4 and SAS/STAT version 15.1 for Windows.

The primary analysis was conducted in the full analysis set, which included all randomized patients, but excluded 3 patients for whom no follow-up data could be obtained and 1 control patient who used off-trial TXA [12]. A sensitivity analysis was performed whereby the primary endpoint for these patients was imputed according to a worst-case scenario (i.e., occurrence of oral bleeding), but yielded similar results. A further sensitivity analysis was performed on the per-protocol set, additionally excluding patients who had not complied with the assigned treatment for at least 80%.

The primary outcome (i.e., number of patients with any oral bleeding) was analyzed using a chi-squared test. The effect of treatment was estimated by the risk ratio and presented along with its 95% confidence interval (CI).

The secondary outcomes were analyzed, and a prespecified subgroup analysis was performed, by means of a logistic regression (for the number of patients with bleeds) and negative binomial regression model (for the number of bleeds). The CIs from the negative binomial regression were calculated using the normal approximation. The treatment effect was estimated as a rate ratio (the number of bleeds per patient during the first 7 days after dental extraction). The subgroup analysis included a risk analysis for bleeding events accounting for patient demographics and procedural characteristics, stratified per treatment group. Additionally, a post hoc exploratory analysis was performed, analyzing all oral bleeds plus facial hematomas by means of a logistic regression (for the primary endpoint) and negative binomial regression model (for the number of bleeds).

Importantly, an independent Data Safety Monitoring Board performed an unplanned interim analysis on the number of patients with oral bleeding in May 2020 with the aim of assessing futility or reassessing the sample size. At that time, the trial was suspended because of the outbreak of the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), and 222 of the 236 planned patients had been randomized. The Data Safety Monitoring Board recommended not to reinitiate the study based on the number of patients with bleeding. The study team remained blinded to all study results. The full details of the methodology of the interim and full analysis are outlined in S1 SAP Interim and S2 SAP Final, respectively.

The study flowchart is shown in Fig 1. The first patient was enrolled in February 2018, and the trial was closed in May 2020. Of 293 patients referred for dental extraction, 222 eligible patients were randomized: 108 patients were assigned to TXA mouthwash and 114 to placebo. Three patients were lost to follow-up, and 1 patient took off-trial TXA, so that 106 TXA-treated patients and 112 placebo-treated patients were included in the full analysis set.

The baseline characteristics of the patients are shown in Table 1. The 2 groups were well balanced, except for a difference in sex. The mean age was 74.8 (±8.8) years in the TXA group and 72.7 (±10.7) years in the placebo group. Details of the dental extractions are presented in S1 Table.

The primary and secondary bleeding outcomes are summarized in Table 2. There was no difference in the number of patients with any oral post-extraction bleeding in the TXA group (28 patients, 26.4%) versus the placebo group (32 patients, 28.6%) (relative risk, 0.92; 95% CI, 0.60 to 1.42; P = 0.72). When assessing all oral bleeding events, there were 46 oral bleeds in the TXA group versus 85 in the placebo group (rate ratio, 0.57; 95% CI, 0.31 to 1.04, P = 0.07). The procedural bleeding score was similar between both groups: 4 ± 1.78 on a VAS scale (0 to 10) for TXA-treated patients and 4 ± 1.82 for placebo-treated patients (P = 0.80).

Patients in the TXA group had less delayed oral bleeds than patients in the placebo group (11 versus 36 delayed bleeds, rate ratio, 0.32; 95% CI, 0.12 to 0.89). TXA-treated patients had fewer bleeds related to the dental extraction, which were oral bleeds and facial hematomas, compared to patients assigned to placebo (52 versus 99 bleeds, rate ratio, 0.56; 95% CI, 0.32 to 0.96).

All oral bleeds in this trial were managed with local hemostatic measures, except for one major oral bleeding in a patient in the placebo arm who needed a blood transfusion and who was hospitalized for 2 days.

Of the 60 patients who reported oral bleeding, 28 experienced more than 1 bleeding (Fig 2). No patient had 4 or more oral bleeds in the TXA group, whereas 10 patients in the placebo group had. Fig 3 shows the number of oral bleeds per day after dental extraction. TXA had no influence on early post-extraction oral bleeding (day 0 and day 1) (see also Table 2). In contrast, the number of oral bleeds was lower on days 2, 3, and 4 after extraction for patients in the TXA group compared to the placebo group (see also Table 2).

Fewer patients treated with TXA (6.6%) had an unplanned medical contact after dental extraction compared to patients using placebo (16.1%) (relative risk, 0.41; 95% CI, 0.18 to 0.94) (Table 3). The most frequent reasons for unplanned medical contact were bleeding and infection (S2 Table). The number of patients requiring a reintervention for oral bleeding or that interrupted their NOAC was comparable between both groups (Table 3). The majority of unplanned NOAC interruptions was because of bleeding complications (S2 Table).

Only 1 patient included in the placebo group suffered from a thrombotic event in preparation for dental extraction (Table 3). The participant had stopped NOAC treatment 2 days before dental extraction and had a transient ischemic attack. There were no thrombotic events in the TXA group. Allergic reactions were rare. Two patients in both treatment groups reported a tingling feeling in their mouth after using the mouthwash (Table 3).

The exploratory subgroup analysis is shown in Fig 4. A significant treatment interaction was identified in patients requiring extraction of 2 or more teeth. In this subgroup, there were 33 bleeds in 19/61 (31.2%) patients assigned to TXA and 77 bleeds in 24/57 (42.2%) of patients assigned to placebo (rate ratio, 0.40; 95% CI, 0.20 to 0.78). In patients who were 75 years or older, there were 24 bleeds in 13/59 (22.0%) patients assigned to TXA and 55 bleeds in 16/51 (31.4%) patients assigned to placebo (rate ratio, 0.35; 95% CI, 0.15 to 0.79), although this interaction did not reach statistical significance (P = 0.07).