Generalized propensity score for estimating the average treatment effect of multiple treatments

Abstract

The propensity score method is widely used in clinical studies to estimate the effect of a treatment with two levels on patient's outcomes. However, due to the complexity of many diseases, an effective treatment often involves multiple components. For example, in the practice of Traditional Chinese Medicine (TCM), an effective treatment may include multiple components, e.g. Chinese herbs, acupuncture, and massage therapy. In clinical trials involving TCM, patients could be randomly assigned to either the treatment or control group, but they or their doctors may make different choices about which treatment component to use. As a result, treatment components are not randomly assigned. Rosenbaum and Rubin proposed the propensity score method for binary treatments, and Imbens extended their work to multiple treatments. These authors defined the generalized propensity score as the conditional probability of receiving a particular level of the treatment given the pre-treatment variables. In the present work, we adopted this approach and developed a statistical methodology based on the generalized propensity score in order to estimate treatment effects in the case of multiple treatments. Two methods were discussed and compared: propensity score regression adjustment and propensity score weighting. We used these methods to assess the relative effectiveness of individual treatments in the multiple-treatment IMPACT clinical trial. The results reveal that both methods perform well when the sample size is moderate or large.

Clinical features of colorectal mucinous and poorly differentiated adenocarcinomas; study concept of a propensity score analysis in a pooled data of 5530 patients

“…We used the propensity score weighting to balance more strictly in essential variables for the comparison analyses that followed 6) . Propensity scores were estimated using a multinomial logistic regression model including age, sex, tumor locations, extent of lymphadenectomy, and pathological disease stage.…”

Recurrent Stroke in Minor Ischemic Stroke or Transient Ischemic Attack With Metabolic Syndrome and/or Diabetes Mellitus

“…All baseline variables were included to calculate the generalized propensity score. Then, HRs with their CIs were estimated by Cox regression models with adjustment of propensity score or weighting of inverse probability of METS/DM category 23. We also performed a similar analysis using METS (IDF) criteria in a sensitivity analysis.…”

Propensity score based methods for estimating the treatment effects based on observational studies.

“…It is also important to make causal inference when multiple groups are involved. Some theoretical results on this area have been established recently (Imbens, 2000;Lechner, 2001;Imai and Van Dyk, 2004), and implemented to estimate ATT using GBM (McCaffrey et al, 2013) or using multinomial logistic regression (Feng et al, 2012). A thorough investigation on estimating both ATT and ATE for multiple group comparisons will have a great value.…”

“…In a RCT the subjects are randomly assigned to treatment groups and it is assumed that all confounding baseline covariates either measured or unmeasured are balanced (Austin, 2011), and therefore treatment effect can be directly estimated by the difference of observed group means. However, it is not always feasible to conduct an RCT due to ethics, cost, and patient preferences (Feng et al, 2012). On the other hand, observed data from patients under different treatments in a natural health care setting, which is termed as observational study, can be available.…”

“…On the other hand, McCaffrey et al (2013) used GBM to improve the quality of the estimate of the unknown generalized propensity score. It has been seen that only multinomial logistic regression (Feng et al, 2012;Imbens, 2000) and GBM (McCaffrey et al, 2013) are used to estimate the generalized propensity score. In this project, an adaptive optimal ensemble method (Datta et al, 2010) is developed to estimate the generalized propensity score by ranking the performances of multinomial logistic regression, random forests, and GBM in balancing covariates.…”