The Risk for Fatal Pulmonary Embolism after Discontinuing Anticoagulant Therapy for Venous Thromboembolism

Context The decision to discontinue anticoagulation for venous thromboembolism (VTE) should depend in part on the subsequent risk for a fatal recurrence. Previous studies have measured fatality rates in patients still receiving anticoagulant drugs. Contribution In a cohort from 2 source studies that had monitored 2052 patients for an average of 4.5 years after discontinuing anticoagulation for a first episode of VTE, the annual risk for any fatal pulmonary embolism (PE) was 0.43 events per 100 patient-years. The risk for definite or probable fatal recurrent PE was 0.17 events per 100 patient-years. Implication Although these rates seem low, decision makers must also take into account the rates of fatal PE recurrence and fatal hemorrhage if anticoagulant therapy is continued. The Editors When deciding whether to discontinue anticoagulant therapy for venous thromboembolism (VTE), the subsequent risk for fatal pulmonary embolism (PE) is an important prognostic consideration. This risk can be expressed in absolute terms as the annual risk for fatal PE and in conditional terms as the risk that recurrent disease will be fatal (case-fatality rate) (1, 2). Knowledge of the annual risk for fatal PE may influence decisions about discontinuing anticoagulation, depending on the comfort level of clinicians and patients about an acceptably low risk for fatal PE. The case-fatality rate measures the clinical impact of disease recurrence if anticoagulation is discontinued, which can be weighed against the case-fatality rate of bleeding if anticoagulation is continued (3, 4). Past studies do not adequately address these issues. They include meta-analyses, which are retrospective and potentially unreliable (2); patient registry or population-based cohort studies that did not standardize anticoagulant therapy (59); or studies from linked administrative databases, which rely on secondary data sources that may be inaccurate (1012). Furthermore, these studies did not monitor patients for recurrent disease for long enough (usually 1 year) to assess long-term risk for fatal PE. Finally, these studies provide limited data on the risk for fatal PE after anticoagulation is discontinued in patient subgroups, such as persons presenting with PE or idiopathic VTE (2, 512). Therefore, we studied an inception cohort of patients who, after a first episode of VTE, had discontinued anticoagulant therapy and underwent clinical follow-up. We created the inception cohort by pooling data from 2 source studies (1315). Our objective was to provide reliable and precise estimates of the annual risk for fatal PE and the case-fatality rate of recurrent PE. We also assessed these outcomes according to patients' initial presentation (deep venous thrombosis [DVT], PE, or both) and VTE etiology (secondary or idiopathic). Methods Inception Cohort We studied an inception cohort of patients who completed at least 3 months of anticoagulant therapy after a first episode of symptomatic VTE. The observation period for the inception cohort began when the patient discontinued anticoagulant therapy, and it ended when the patient had nonfatal DVT or PE, fatal PE, or death from another cause. If 1 of these outcomes did not occur, patients were observed until the study ended or for a maximum of 120 months, whichever came first. Inception Cohort Subgroups We placed patients in subgroups according to initial presentation of VTE (DVT alone, PE alone, or DVT and PE) and VTE etiology (secondary or idiopathic). We defined secondary VTE as occurring in the presence of 1 or more transient risk factors, such as surgery, leg trauma, leg fracture, or childbirth (within 3 months of presentation), or if the patient was bedridden for more than 1 week because of a medical illness, rheumatologic illness, pregnancy, or hormonal therapy use. We defined idiopathic VTE as occurring in the absence of any of the aforementioned risk factors. Clinical Outcomes During follow-up of the inception cohort, we assessed the following outcomes: recurrent nonfatal DVT (defined as a new noncompressible vein segment, an increase of the vein diameter by >4 mm compared with the last available measurement on venous ultrasonography, or a new intraluminal filling defect on venography); recurrent nonfatal PE (defined as a new ventilationperfusion mismatch on lung scan or a new intraluminal filling defect on spiral computed tomography of the chest); definite or probable fatal PE; and possible fatal PE (sudden death of undetermined cause). We determined a cause of death for all deceased patients on the basis of autopsy records, hospital death certificates, and national death registries. The diagnosis of fatal PE may be inaccurate because of bias from diagnostic suspicion about the cause of death in patients with previous VTE (16, 17). Therefore, we attempted to provide a more robust assessment of the annual risk for fatal PE by classifying fatal PE as definite or probable fatal PE or as any fatal PE (definite or probable PE and possible PE). A fatal PE was adjudicated as definite if it was confirmed at autopsy, probable if it was preceded in the immediate period before death by confirmed nonfatal PE or DVT, and possible if sudden death occurred that could not be explained by a disease or a condition other than PE. Patients who had recurrent nonfatal VTE did not have further follow-up as part of our inception cohort because these patients would typically resume anticoagulation, which would make them ineligible for the inception cohort, and we had defined this cohort as no longer receiving anticoagulation. Source Studies for Derivation of the Inception Cohort The inception cohort was derived from the pooled patient populations of 2 source studies: a prospective cohort study of patients with VTE who received at least 3 months of anticoagulant therapy (13) and an open-label randomized trial that compared 6 weeks and 6 months of anticoagulant therapy in patients with VTE (14, 15). We included patients from the randomized trial only if they had been in the study arm that was randomly assigned to receive 6 months of anticoagulation because 6 weeks is less than the recommended duration of treatment for VTE (18). Although the inception cohort was derived from 2 patient populations, the source studies had common characteristics that allowed pooling of the populations into 1 homogeneous inception cohort. We describe these characteristics, common to both source studies, below. Patient Characteristics In both studies, the study population consisted of consecutive patients presenting with a first episode of symptomatic lower-limb DVT, PE, or both (patients with asymptomatic VTE were excluded) who were referred by a primary care physician (from a family practice or emergency department). Both studies excluded patients if they had at least 1 of the following conditions that would typically warrant lifelong anticoagulation: active cancer, permanent immobility, or high-risk thrombophilia. Diagnosis of DVT and PE Deep venous thrombosis and PE were confirmed by objective diagnostic testing in all patients in both studies. In the cohort study, all patients with DVT had venous ultrasonographyconfirmed proximal DVT, and proximal DVT was defined as DVT involving the popliteal or more proximal deep vein segments. In the randomized trial, patients with DVT had venography-confirmed proximal or distal (calf) DVT, depending on whether the thrombus was proximal or distal to the knee joint. Thus, patients with DVT in the popliteal vein (all of whom would have been classified as having proximal DVT by the cohort study criteria) had proximal DVT if the thrombosed vein segment was above the knee joint and distal DVT if the thrombosed vein segment was below the knee joint. Anticoagulant Therapy All patients in both studies received 5 to 10 days of unfractionated or low-molecular-weight heparin followed by a vitamin K antagonist, administered to achieve an international normalized ratio of 2.0 to 3.0. In the cohort study, therapy was given for at least 3 months but was left to the discretion of the treating physician in all other regards. Patients with secondary VTE typically received 3 to 6 months of anticoagulation, and patients with idiopathic VTE typically received 6 months or more of treatment. In the randomized trial, anticoagulant therapy was given for 6 weeks or 6 months, but we included only patients allocated to receive 6 months of treatment. Follow-up after Discontinuing Anticoagulant Therapy In both source studies, the authors followed patients for a prespecified period unless a patient died or was lost to follow-up. The length of patient follow-up differed according to the source study. In the cohort study, the follow-up protocol was a clinic visit or telephone call every 6 months after stopping anticoagulation until the study was completed or for a maximum of 120 months after discontinuation of anticoagulant therapy. The protocol for the randomized trial was a clinic visit at 3 and 6 months during the first year after discontinuation of anticoagulant therapy, and every 12 months thereafter for a maximum of 120 months after anticoagulant therapy was discontinued. Assessment of Clinical Outcomes Both source studies had independent adjudication committees that reviewed all clinical outcomes. The adjudication committee members in both source studies were unaware of patients' clinical characteristics at presentation or their treatment. Statistical Analysis We obtained the data sets from both source studies and combined them into a single data set so that we could perform a patient-level analysis of the outcomes of interest. We used mean values and ranges to describe the clinical characteristics of patients at the time of initial presentation, when DVT or PE was diagnosed. We characterized anticoagulant therapy as the duration of treatment in months (mean and range) and as the proportion of patients who received therapy for 3 months, more th