Methods/Statistics
The Complex Estimand of Clone-Censor-Weighting When Studying Treatment Initiation Windows Quoa Her* Michael Webster-Clark Yi Li Sophie Dell’Aniello Robert Platt
Clone-censor-weighting (CCW) helps researchers compare treatment regimens that are indistinguishable from one another at baseline without relying on landmark dates or creating immortal person time. One application of the design in observational data is comparing outcomes of starting treatment in specific time windows (e.g., starting inhaled corticosteroids 30 days or from 30 to 180 days following hospitalization for chronic obstructive pulmonary disease exacerbation). Unfortunately, the hypothetical intervention that underlies the quantity that CCW estimates in such cases is more complex than it may seem.
We show that using CCW to study a regimen such as “start treatment prior to day 30” estimates the potential outcome of a hypothetical intervention where A) prior to day 30, everyone follows the treatment start distribution of the study population and B) everyone who has not initiated by day 30 initiates on day 30. Figure 1 illustrates how the resulting cumulative initiation probability would vary for four populations following different patterns of treatment initiation. We further show that allowing all initiators prior to day 30 to contribute to inverse probability of censoring weights (IPCW) estimates the effect of an impossible intervention where those who are forced to start on day 30 have the exposure history of those who started prior, creating bias. This bias is removed if only those who initiate on day 30 contribute to IPCW.
Finally, we demonstrate these findings via a simple simulation and show that, in some special cases, this treatment effect collapses to something more interpretable and those who initiate prior to day 30 can contribute to IPCW. While CCW is an excellent way to estimate outcomes of interventions such as “start by 30 days,” providing information on the timing of treatment initiation and the plausibility of cumulative treatment effects generally provides key context on the underlying intervention.
