The Comparison of ACI and MCB Methods for Choosing a Set that Contains the Optimal Dynamic Treatment Regime

Dynamic treatment regimes (DTRs) are used to operationalize treatment decision-making at different stages by clinicians. The decision rules are based on the time-varying patient characteristics. DTRs can provide more effective decisions compared with once-and-for-all decisions. Here we compare two approaches to identify a set of DTRs that includes the optimal DTR. The methods are: the ACI (Adaptive Confidence Intervals) method by Laber et al. with our modifications and the MCB (Multiple Comparisons with the Best) method by Ertefaie et al. We simulate data from four different scenarios to compare the MCB method and a modified version of the ACI method. By comparing the probabilities that the best DTR is included into the constructed set, and the average set size of each method in four different scenarios, we find that both methods include the true optimal DTR with a specified probability. The MCB method generally has a smaller average set size, indicating that it has more power in excluding inferior DTRs. Thus, we conclude that MCB method performs better in general. Keyword: DTR, SMART, multiple comparison