Estimating average treatment effects when treatment data are absent in a target study