Estimating hypothetical estimands with causal inference and missing data estimators in a diabetes trial