Propensity score matching and inverse probability of treatment weighting to address confounding by indication in comparative effectiveness research of oral anticoagulants

Abstract

After decades of warfarin being the only oral anticoagulant (OAC) widely available for stroke prevention in atrial fibrillation, four direct OACs (apixaban, dabigatran, edoxaban and rivaroxaban) were approved after demonstrating noninferior efficacy and safety versus warfarin in randomized controlled trials. Comparative effectiveness research of OACs based on real-world data provides complementary information to randomized controlled trials. Propensity score matching and inverse probability of treatment weighting are increasingly popular methods used to address confounding by indication potentially arising in comparative effectiveness research due to a lack of randomization in treatment assignment. This review describes the fundamentals of propensity score matching and inverse probability of treatment weighting, appraises differences between them and presents applied examples to elevate understanding of these methods within the atrial fibrillation field.

Electroconvulsive therapy is associated with lower readmission rates in patients with schizophrenia

“…With the development of machine learning algorithms, screening for critical risk factors that affect the disease course can be applied to medical diseases including psychiatric disease [20][21][22] . The XGBoost model achieves the best prediction modelling among the machine learning models and traditional regression models [23] .…”

Impact of Chronic Obstruction Pulmonary Disease with Acute Exacerbation on Survival in Current-Smoking Patients with Advanced stages Lung Squamous Cell Carcinoma undergoing Concurrent Chemoradiotherapy

“…[10,[46][47][48][49] Although matching creates a balanced dataset by making pairs between controls and treated patients on the basis of a similar propensity score, some patients may be excluded from the dataset, which is a major disadvantage. [50,51] In the current study, IPTW (Table 3) was conducted, which has advantages over matching of patients based on propensity scores when there are two groups to compare, when nding matches results in insu cient sample sizes, or when the data are censored. [50,51] Moreover, the data in Table 1 are real-world data; thus, we did not use propensity score matching to wash out too much sample of patients induced a deviation database.…”

“…[50,51] In the current study, IPTW (Table 3) was conducted, which has advantages over matching of patients based on propensity scores when there are two groups to compare, when nding matches results in insu cient sample sizes, or when the data are censored. [50,51] Moreover, the data in Table 1 are real-world data; thus, we did not use propensity score matching to wash out too much sample of patients induced a deviation database. [50,51] To create a pseudo-study cohort,…”

“…[50,51] Moreover, the data in Table 1 are real-world data; thus, we did not use propensity score matching to wash out too much sample of patients induced a deviation database. [50,51] To create a pseudo-study cohort,…”

Introducing Volume 10 of the Journal of Comparative Effectiveness Research

“…Atrial fibrillation is a type of cardiac arrhythmia that affects millions worldwide, and patients are often prescribed oral anticoagulants to reduce the risk of stroke. In their review article published in issue 9, Allan et al describe the fundamentals of propensity score matching and inverse probability of treatment weighting methods to address confounding by indication in real-world studies, appraise similarities and differences between these techniques and give illustrative examples from some case studies comparing the effectiveness and safety of oral anticoagulants [6].…”