Hi, all! I was hoping I can get some insight on a critique I received from a peer reviewer. For context, I needed to balance my intervention and comparison groups as the study used a quasi-experimental design. I did so by estimating propensity scores on the relevant baseline covariates and then applied inverse probability of treatment weights (IPTW) calculated with the average treatment effect on the treated estimand.
For my outcome models, I estimated GLMs that included the binary indicator for study group, baseline of the outcome, sex, age, and race (with cases weighted by the IPTW). Baseline of the outcome, sex, age, and race were included in the propensity score model. I did this because sex, age, and race were pre-specified biological control variables in my NIH grant, and including the baseline outcome also made substantive and methodological sense.
A peer reviewer questioned my choice to control for those covariates in the outcome models since they were also included in the propensity score model and said that I should acknowledge this as a limitation. They did not cite any literature to support this critique, and I can only find one article that cautions against doing so but specifically in survival analysis (so not relevant) or controlling for baseline covariates that are not included in the propensity score models (not the case).
Particularly given that the foundation of doubly robust estimation is IPTW + covariate adjustments, I do not see this choice warranting a limitation, but perhaps I am missing something? I was curious if anybody has any insight on this? I don't want to push back if this is indeed a valid critique/limitation, but I also don't want to note it as a limitation if that is not the case.