Estimation of current leukaemia-free survival following donor lymphocyte infusion therapy for patients with leukaemia who relapse after allografting: application of a multistate model.

Allogeneic bone marrow transplantation has become routine treatment for selected patients with leukaemia and other haematological disorders. A standard measure of the efficacy of this treatment for patients is assessment of leukaemia-free survival, that is, the probability of being alive and in remission at different time intervals after transplant. During the last 10 years a new approach to managing patients who relapse after allografting was developed - namely infusion of lymphocytes (white blood cells) collected from the original donor. Such donor lymphocyte infusions (DLI) are highly effective in restoring complete remission and it appears that these remissions are durable. Consequently estimates of long-term success of allografting must consider salvage of initial relapses with DLI. To do so a new estimator of treatment efficacy, called 'current leukaemia-free survival' is sometimes used in the medical literature. This curve attempts to estimate the probability that a patient is alive in an original remission or in subsequent remission after treatment with DLI at a given time after transplant. Here we show that this estimator does not actually estimate the probability of interest and we use a multi-stage model to develop a new and more appropriate estimator for 'current leukaemia-free survival'.