Handling missing data when estimating causal effects with targeted maximum likelihood estimation | Synapse