Should a propensity score model be super? The utility of ensemble procedures for causal adjustment.

In investigations of the effect of treatment on outcome, the propensity score is a tool to eliminate imbalance in the distribution of confounding variables between treatment groups. Recent work has suggested that Super Learner, an ensemble method, outperforms logistic regression in nonlinear settings; however, experience with real-data analyses tends to show overfitting of the propensity score model using this approach. We investigated a wide range of simulated settings of varying complexities including simulations based on real data to compare the performances of logistic regression, generalized boosted models, and Super Learner in providing balance and for estimating the average treatment effect via propensity score regression, propensity score matching, and inverse probability of treatment weighting. We found that Super Learner and logistic regression are comparable in terms of covariate balance, bias, and mean squared error (MSE); however, Super Learner is computationally very expensive thus leaving no clear advantage to the more complex approach. Propensity scores estimated by generalized boosted models were inferior to the other two estimation approaches. We also found that propensity score regression adjustment was superior to either matching or inverse weighting when the form of the dependence on the treatment on the outcome is correctly specified.

A cure-rate model for Q-learning: Estimating an adaptive immunosuppressant treatment strategy for allogeneic hematopoietic cell transplant patients.

Cancers treated by transplantation are often curative, but immunosuppressive drugs are required to prevent and (if needed) to treat graft-versus-host disease. Estimation of an optimal adaptive treatment strategy when treatment at either one of two stages of treatment may lead to a cure has not yet been considered. Using a sample of 9563 patients treated for blood and bone cancers by allogeneic hematopoietic cell transplantation drawn from the Center for Blood and Marrow Transplant Research database, we provide a case study of a novel approach to Q-learning for survival data in the presence of a potentially curative treatment, and demonstrate the results differ substantially from an implementation of Q-learning that fails to account for the cure-rate.

Non-HIV Vaccine-Induced Immune Responses as Potential Baseline Immunogenicity Predictors of ALVAC-HIV and AIDSVAX B/E-Induced Immune Responses.

Identifying correlations between immune responses elicited via HIV and non-HIV vaccines could aid the search for correlates of HIV protection and increase statistical power in HIV vaccine-efficacy trial designs. An exploratory objective of the HVTN 097 phase 1b trial was to assess whether immune responses [focusing on those supported as correlates of risk (CoR) of HIV acquisition] induced via the RV144 pox-prime HIV vaccine regimen correlated with those induced via tetanus toxoid (TT) and/or hepatitis B virus (HBV) vaccines. We measured TT-specific and HBV-specific IgG-binding antibody responses and TT-specific and HBV-specific CD4+ T-cell responses at multiple time points in HVTN 097 participants, and we assessed their correlations at peak time points with HIV vaccine (ALVAC-HIV and AIDSVAX B/E)-induced responses. Four correlations were significant [false discovery rate-adjusted p-value (FDR) ≤ 0.2]. Three of these four were with IgG-binding antibody responses to TT measured one month after TT receipt, with the strongest and most significant correlation [rho = 0.368 (95% CI: 0.096, 0.588; p = 0.008; FDR = 0.137)] being with IgG-binding antibody responses to MN gp120 gDneg (B protein boost) measured two weeks after the second ALVAC-HIV and AIDSVAX B/E boost. The fourth significant correlation [(rho = 0.361; 95% CI: 0.049, 0.609; p = 0.021; FDR = 0.137)] was between CD4+ T-cell responses to a hepatitis B surface antigen peptide pool, measured 2 weeks after the third HBV vaccination, and IgG-binding antibody responses to gp70BCaseAV1V2 (B V1V2 immune correlate), measured two weeks after the second ALVAC-HIV and AIDSVAX B/E boost. These moderate correlations imply that either vaccine, TT or HBV, could potentially provide a moderately useful immunogenicity predictor for the ALVAC-HIV and AIDSVAX B/E HIV vaccine regimen.