gms | German Medical Science

Studying paediatric leukaemia patients with and without available stem cell donors enables to compare the efficacy of stem cell transplantation with chemotherapy. Since donor search is expensive and takes time, a patient's donor availability status becomes known either when a donor is identified or when no donor is found after extensive searching.Donor availability remains unknown when donor search is cancelled for both cost and ethical reasons. This usually happens in case of premature death and when stem cell transplantation is no longer considered a suitable treatment option (e.g. after a relapse). Technically, donor availability status can be considered as a partly unobserved, external, binary time-dependent covariate. The statistical situation becomes even more challenging due to non-proportional hazards resulting from an increased post-transplantation mortality that vanishes over time.

The generalised pseudo-values approach is able to correctly address these issues and enable an unbiased treatment comparison. However, calculations are time-consuming due to a necessary bootstrap step. A faster alternative approach is suggested which is based on weighting of common pseudo-values and does not require resampling.

Both approaches are compared by a simulation study where they show similarly satisfactory behaviour with respect to confidence interval coverage and type I errors. A real data example is presented and analysed to illustrate and discuss both approaches.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.