Adjusted curves for clustered survival and competing risks data

Observational studies with right-censored data often have clustered data due to matched pairs or a study center effect. In such data, there may be an imbalance in patient characteristics between treatment groups, where Kaplan-Meier curves or unadjusted cumulative incidence curves can be misleading and may not represent the average patient on a given treatment arm. Adjusted curves are desirable to appropriately display survival or cumulative incidence curves in this case. We propose methods for estimating the adjusted survival and cumulative incidence probabilities for clustered right-censored data. For the competing risks outcome, we allow both covariate-independent and covariate-dependent censoring. We develop an R package adjSURVCI to implement the proposed methods. It provides the estimates of adjusted survival and cumulative incidence probabilities along with their standard errors. Our simulation results show that the adjusted survival and cumulative incidence estimates of the proposed method are unbiased with approximate 95% coverage rates. We apply the proposed method to stem cell transplant data of leukemia patients.