Optimal length of oral anticoagulant treatment for maximum benefit within 5 years after discontinuation of oral anticoagulants in patients with acute pulmonary embolism who require secondary thromboprophylaxis

Study design

A retrospective study was performed to explore the optimal length of OA treatment for maximum benefit within 5 years after OA discontinuation in patients with acute PE who required secondary thromboprophylaxis. Patients were reviewed if they had an objectively confirmed PE and underwent extended OA treatment for secondary thromboprophylaxis after the completion of at least 6 months of OA treatment following the diagnosis of acute PE and discontinued OA use at least 5 years prior to the current study. The primary outcome was the mortality rates, including all-cause and PE-related mortality within 5 years after OA discontinuation. PE-related mortality was defined as those in which PE or VTE was listed as the underlying cause of death.9 The secondary outcomes comprised VTE recurrence, major bleeding during OA use and net clinical benefit. VTE recurrence was defined as the incidence of recurrent VTE within 5 years after OA discontinuation. Major bleeding during OA use was defined as the incidence of major bleeding during the OA treatment. Net clinical benefit was defined as the composite of recurrent VTE and major bleeding.10 To be specific, it referred to patients with neither recurrent VTE within 5 years after the OA discontinuation nor major bleeding during OA use.

VTE recurrence was defined as the presence of new defect on ventilation/perfusion (V/Q) scan or new thrombi on CT pulmonary angiography or newly developed thrombi on compression ultrasonography, in the re-examination during follow-up.11 The determination of VTE recurrence via imaging investigation was conducted in the context of patients having sudden aggravation of VTE symptoms or routine follow-up examination every 3 or 6 months. Major bleeding was defined as fatal bleeding, symptomatic bleeding in critical areas or organs, surgical site bleeding causing second intervention and/or haemodynamic instability, haemarthrosis, resulting in immobilisation, prolonged hospitalisation, delayed wound healing or deep wound infection, and bleeding causing a fall in haemoglobin level of 20 g/L or more, or a transfusion of two or more units of whole blood or red cells with temporal association within 24–48 hours to the bleeding.12 13

The post hoc analysis was conducted when the required data were completely collected. According to the overall length of OA use that patients had undergone after the diagnosis of acute PE, they were classified into three different groups, which were group 1 (≥1 year of OA use and <2 years of OA use), group 2 (≥2 years of OA use and <3 years of OA use) and group 3 (≥3 years of OA use). We used propensity score-matching methods to abate the bias of potential confounding factors highly related to mortality, VTE recurrence and bleeding in the ESC (European Society of Cardiology) guidelines,4 which mainly included simplified pulmonary embolism severity index, risk strata of PE, D-dimer, age, sex, hormonal therapy (DASH) score and hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalised ratio, elderly (>65 years), drugs/alcohol concomitantly (HAS-BLED) score. The types of OA and cancer were also incorporated into the propensity score matching. Mortality, VTE recurrence, major bleeding and net clinical benefit were compared among different groups, respectively. The correlation between the length of OA use and mortality, VTE recurrence as well as major bleeding were analysed. Different length of OA use for the tendency of mortality, VTE recurrence, bleeding and net clinical benefit were analysed. The time-dependent incidence of mortality, VTE recurrence and major bleeding were compared among three groups with different length of OA use.

The investigators of Shanghai Xinhua Hospital, Shanghai Pulmonary Hospital and Shanghai Punan Hospital performed the current study. The data needed for study were obtained mainly through the existing electronic medical record system or through telephone communication. No one who is not an author contributes to the writing of manuscript. All authors vouch for the completeness, fidelity and accuracy of the data. All authors have read and approved the submitted version of manuscript for publication. The study protocol was approved by the institutional review board of each participating hospital. Due to the nature of retrospective study, written informed consent for each patient was waived by the institutional review board of each participating hospital. For patients being contacted through telephone communication, verbal consent was all obtained.

Study population

Patients were incorporated according to the inclusion and exclusion criteria. The inclusion criteria consisted of: (1) all patients were 18 years old or older when they were diagnosed with acute PE; (2) all patients had an objectively confirmed PE and underwent extended OA treatment for secondary thromboprophylaxis due to the high risk of recurrence and low risk of bleeding after the completion of at least 6 months of OA anticoagulation following the diagnosis of acute PE; (3) all patients had discontinued OA use for at least 5 years prior to the current study. The exclusion criteria consisted of: (1) patients who already had a definite chronic thromboembolic disease (CTED) history prior to the acute PE onset; (2) patients who underwent a concurrent use of single or dual antiplatelet therapies during OA use; (3) patients whose OA use were interrupted for more than 3 months during the entire anticoagulation course; (4) patients who resumed OA use after discontinuation.

Statistical analysis

In the propensity score-matching analysis, nearest-neighbour (greedy) matching without replacement was adopted.14 Since the longer the OA use, the fewer the number of patients, we used a 4:2:1 ratio within a calliper width of 0.215 of the SD of the logit of the propensity score for the number ratio of patients in groups 1, 2 and 3. The matching algorithm first selected a patient in group 3 and then selected two patients in group 2 and 4 patients in group 1 who had a linear propensity score that were closest to that of the first selected patient.

The comparison of measurement data was performed by using t test or ANOVA (Analysis of Variance). The between-group comparison of rates was performed by χ² test. The correlation between the length of OA use and mortality, VTE recurrence as well as major bleeding were analysed by using Cox regression analysis. Different length of OA use for the tendency of mortality, VTE recurrence, bleeding and net clinical benefit were analysed by using receiver operator characteristic (ROC) curve analysis. The time-dependent incidence of mortality, VTE recurrence (the first episode) and major bleeding (the first episode) were compared among groups with different length of OA use by using Kaplan-Meier curve analysis. The statistical analyses were performed by using SPSS V.26 and R software, V.3.6.1 (R Project for Statistical Computing). A p value being less than 0.05 denoted statistical significance.

Demographics and characteristics of patients at the PE diagnosis

A total of 606 patients during the period between 2012 and 2022 were incorporated from the participating hospitals as per the inclusion criteria. Then as per the exclusion criteria, 94 patients were excluded from the current study. As such, 512 patients were acquired via the inclusion and exclusion criteria. After the following propensity-score matching, a total of 385 patients with PE entered into the final analysis at last. The median age of all patients was 69.0 years old. Among them, 177 patients were women, whereas 208 patients were men. The distribution of all 385 patients in the three groups was 220 in group 1 (≥1 year of OA use and <2 years of OA use), 110 in group 2 (≥2 year of OA use and <3 years of OA use) and 55 in group 3 (≥3 years of OA use), respectively. The types of OA included warfarin, dabigatran and rivaroxaban. The dose of extended warfarin was still adjusted to maintain an international normalised ratio to be 2 to 3, whereas that of almost all dabigatran (92.7%) was 110 mg two times per day, and of almost all rivaroxaban (92.5%) was 10 mg/day. Almost all patients continued to use the OA type which was used for the initial 6 months in the extended anticoagulation, whereas only 23 (6.0%) patients ever changed from one type of OA to another during the entire anticoagulation treatment course. The median OA treatment duration was 2.8 (1.1–5.5) years. The main reasons for extended OA use principally include active cancers, previous VTE history and hypercoagulability such as inflammatory bowel disease, antiphospholipid syndrome, rheumatoid arthritis and so on. The main reasons for OA discontinuation principally include empirical advice for OA discontinuance from doctors, the occurrence of adverse events and requested withdrawal by patients. The demographics and characteristics of patients are summarised in table 1.

Outcomes among groups with different length of OA use

The overall number of cumulative all-cause mortality, PE-related mortality, VTE recurrence, major bleeding and net clinical benefit were 135 (35.1%), 20 (5.2%), 111 (28.8%), 15 (3.9%) and 259 (67.3%), in all 385 patients. The number of cumulative all-cause mortality 5 years after the OA discontinuation was 80 (36.4%), 37 (33.6%) and 18 (32.7%), in the group 1, 2 and 3, respectively (p=0.821). No statistical difference existed in pairwise comparison (all p>0.05). Among all deceased patients, the number of PE-related mortality was 14(6.4%), 4(3.6%) and 2(3.6%), in the group 1, 2 and 3, respectively (p=0.491). Nevertheless, a pairwise comparison demonstrated that the PE-related mortality in group 1 was higher than that in group 2 (p=0.034) and 3 (p=0.040), respectively, whereas were similar between groups 2 and 3 (p=1.000). The number of patients who had at least one time of VTE recurrence were 73 (33.2%), 26 (23.6%) and 12 (21.8%), in the group 1, 2 and 3, respectively (p=0.038). A pairwise comparison showed that the VTE recurrence in group 1 was higher than that in group 2 (p=0.021) and 3 (p=0.035), respectively, whereas no statistical difference existed between group 2 and 3 (p=0.794). The number of patients who had at least one time of major bleeding during OA use were 5 (2.3%), 5 (4.5%) and 5 (9.1%), in the groups 1, 2 and 3, respectively (p=0.025). None of them was fatal bleeding. A pairwise comparison showed that the major bleeding in group 3 was higher than that in groups 1 (p=0.020) and 2 (p=0.029), respectively, whereas no statistical difference existed between group 1 and 2 (p=0.291). The number of patients who had net clinical benefit were 142 (64.5%), 79 (71.8%) and 38 (69.1%), in the groups 1, 2 and 3, respectively (p=0.395). A pairwise comparison showed that the net clinical benefit in group 1 was less than that in group 2 (p=0.024), whereas similar with that in group 3 (p=0.526). The net clinical benefit was comparable between group 2 and 3 (p=0.716). The comparison of outcomes among groups with different length of OA use is demonstrated in figure 1.

Request permission

Correlation between the length of OA use and mortality, VTE recurrence, bleeding as well as net clinical benefit

By incorporating outcome-related factors for PE such as sPESI, risk strata of PE, DASH score and HAS-BLED score, a multivariate Cox regression analysis demonstrated that the length of OA use was positively associated with major bleeding (HR, 2.510 (0.293 to 3.485), p=0.001), whereas negatively associated with PE-related mortality (HR, 0.668 (0.196 to 2.832), p=0.025) and VTE recurrence (HR, 0.694 (0.174–2.300), p=0.036), respectively. The length of OA use was not associated with all-cause mortality (HR, 1.033 (0.588 to 1.815), p=0.854). The correlation between the length of OA use and mortality, VTE recurrence as well as major bleeding is demonstrated in table 2.

Different length of OA use for tendency of mortality, VTE recurrence, bleeding, net clinical benefit

In an ROC curve analysis, the length of OA use had no trend towards the all-cause mortality within 5 years after OA discontinuation. The area under the curve (AUC) was 0.508 (0.444 to 0.572) (p=0.802). Nevertheless, the sensitivity and specificity of the length of OA use for the tendency of PE-related mortality within 5 years after OA discontinuation was 70.2% and 46.2%, respectively. The AUC was 0.654 (0.514 to 0.793) (p=0.029). The cut-off point was 27 months (figure 2A). It means that OA use being longer than 27 months may yield lower PE-related mortality than OA use being shorter than 27 months. The ROC curve analysis also demonstrated that the sensitivity and specificity of the length of OA use for the tendency of VTE recurrence within 5 years after OA discontinuation was 70.1% and 55.4%, respectively. The AUC was 0.587 (0.526 to 0.647) (p=0.006). The cut-off point was 33 months (figure 2B). It means that OA use being longer than 33 months may yield lower incidence of VTE recurrence than OA use being shorter than 33 months. The sensitivity and specificity of the length of OA use for the tendency of major bleeding during the OA medication was 91.7% and 61.9%, respectively. The AUC was 0.771 (0.648 to 0.894) (p=0.001). The cut-off point was 25 months (figure 2C). It means that OA use being longer than 25 months may yield higher incidence of major bleeding than OA use being shorter than 25 months. The sensitivity and specificity of the length of OA use for the tendency of net clinical benefit within 5 years after OA discontinuation was 86.8% and 64.3%, respectively. The AUC was 0.628 (0.565 to 0.690) (p<0.001). The cut-off point was 30 months (figure 2D). It means that OA use being longer than 30 months may yield more net clinical benefit than OA use being shorter than 30 months. Different length of OA use for the tendency of mortality, VTE recurrence, bleeding and net clinical benefit are illustrated in figure 2.

Request permission

Comparison of time-dependent incidence of outcomes among groups with different length of OA use

In a Kaplan-Meier curve analysis, no overall difference of time-dependent cumulative all-cause mortality rates was observed among groups 1, 2 and 3 (p=0.853) (figure 3A). Likewise, no overall difference of time-dependent cumulative PE-related mortality rates existed among all three groups (p=0.469). Nevertheless, PE-related death in group one evidently occurred earlier than that in both of groups 2 and 3 (figure 3B). The Kaplan-Meier curve analysis demonstrated that the time-dependent cumulative VTE recurrence-free rates were different among groups 1, 2 and 3 (p=0.010). The patients in group 1 yielded VTE recurrence earlier than those in group 2, who yielded VTE recurrence earlier than those in group 3 (figure 3C). It also revealed that the time-dependent cumulative major bleeding rates during OA use were similar among group 1, 2 and 3 (p=0.443) (figure 3D). The comparison of time-dependent incidence of outcomes among groups with different length of OA use is demonstrated in figure 3.

Request permission