Announcement
Thank you for attending THSNA 2026. The virtual meeting is now closed.
Thank you for attending THSNA 2026. The virtual meeting is now closed.
Presentation Details
| Incidental Pulmonary Embolism: When Treatment May Be Worse Than the Diagnosis Sophie Samuel, Ada Selina Jutba, Haley Parker, Gregory Pon, Jennifer Cortes. Memorial Hermann Health System, Houston, TX, USA |
Abstract
Background: Incidental pulmonary embolism (iPE) is increasingly identified on imaging performed for non-thrombotic indications. Although guidelines often recommend treating iPE similarly to symptomatic pulmonary embolism (sPE), the evidence base is limited, and patients undergoing imaging for other clinical conditions may be more susceptible to bleeding when therapeutic anticoagulation is initiated. Objective: To compare the probability of in-hospital major bleeding between patients with iPE and those with sPE using Bayesian estimation and causal inference methods, and to evaluate the robustness of this association to measured and unmeasured confounding. Methods: We conducted a multicenter retrospective cohort study of 435 adults hospitalized with radiographically confirmed acute pulmonary embolism at four hospitals between January 2023 and September 2024 (71 iPE, 364 sPE). The primary outcome was a composite of in-hospital major bleeding, defined as gastrointestinal, intracranial, or retroperitoneal hemorrhage, or high-volume transfusion (≥10 units of packed red blood cells within 24 hours or ≥4 units within 1 hour). Bayesian binomial models estimated posterior risks, risk differences, and the posterior probability that bleeding risk was higher in iPE. Bayesian multivariable logistic models, including inverse-probability weighting (IPTW) and doubly robust specifications, estimated posterior odds ratios (ORs). Covariates in the propensity model included age, sex, active cancer, chronic kidney disease (CKD), and heart failure or chronic lung disease. Sensitivity analyses excluded patients with CKD or active cancer. An E-value was calculated to assess robustness to unmeasured confounding. Results: Bayesian pooled analysis estimated a posterior bleeding risk of 8.6% (95% credible interval [CrI] 3.6–16.6%) for iPE and 4.4% (95% CrI 2.6–6.9%) for sPE, corresponding to a risk difference of 4.2% (95% CrI −1.4–12.3%) and a posterior probability of 0.92 for higher bleeding risk in iPE. The site-adjusted binomial model estimated risks of 9.9% and 4.8% for iPE and sPE, respectively, with a risk difference of 5.1% (95% CrI 0.1–17.3%) and a posterior probability of 0.96 that bleeding risk was higher in iPE. In the Bayesian multivariable logistic model, iPE was associated with a posterior OR of 1.76 (95% CrI 0.65–4.76). After IPTW, covariate balance was excellent (standardized mean differences <0.10). The weighted marginal structural model produced a posterior OR of 1.92 (95% CrI 0.54–6.75), and the doubly robust weighted model yielded a posterior OR of 1.83 (95% CrI 0.50–6.69). In this IPTW doubly robust model, CKD was independently associated with major bleeding (posterior OR 3.74, 95% CrI 1.27–11.04). Sensitivity analyses excluding patients with CKD or active cancer yielded posterior ORs of 1.71 and 1.95, respectively. The E-value of 3.6 indicated that only a strong unmeasured confounder could fully explain the observed association. Conclusions: Patients with iPE had a higher probability of in-hospital major bleeding than those with sPE, with consistent findings across Bayesian estimation, IPTW, doubly robust modeling, and sensitivity analyses. iPE appears to represent a clinically fragile subgroup in whom therapeutic anticoagulation should be individualized, particularly in the presence of chronic kidney disease.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.