Abstract No. 665 Association of inferior vena cava filter placement and mortality among in-patients with pulmonary embolism after adjusting for immortal time bias

Abstract

Several large observational studies have suggested short-term mortality benefit with IVC filter placement in patients admitted with acute PE. However, these studies have failed to adjust for immortal time bias, i.e., the time between admission and filter placement during which death can only occur in the non-filter group. We assessed the association between IVC filter placement and inpatient mortality in patients admitted with pulmonary embolism, while adjusting for immortal time bias. Patients admitted with pulmonary embolism and undergoing IVC filter placement were retrospectively identified using ICD-9 diagnosis and procedure codes in the 2014 National Inpatient Sample. The association between IVC filter placement and inpatient mortality was assessed using survey-weighted Cox proportional hazard regression models and adjusted for baseline disease severity. Two regression models were fitted to assess the impact of immortal time bias: (1) IVC filter placement as a binary yes/no variable (implicit immortal time bias) and (2) IVC filter placement as a time-dependent variable (accounting for immortal time bias). There were 179,235 national admissions (52.64% females, mean age 62.1 years) for acute pulmonary embolism in 2014. Among these, 16,390 (9.15%) received IVC filters and 5299 (2.96%) patients died before discharge. All-cause mortality among patients who received IVC filters was 605/16,390 (3.69%), compared with 4694/162,815 (2.88%) without a filter. Naïve regression modelling indicated a protective benefit for IVC filter placement (hazard ratio:0.64, 95% CI: 0.53- 0.77). However, this apparent benefit was mitigated with inclusion of IVC filter placement as time-dependent variable (hazard ratio 0.86, 95% CI: 0.70-1.04). IVC filter placement does not protect against inpatient mortality in patients admitted with acute pulmonary embolism after accounting for immortal time bias. Future observational studies should adjust for immortal time bias when reporting association of mortality with IVC filter placement.