| Popis: |
In the last decade, the number of older patients undergoing allogeneic Hematopoietic Cell Transplantation (alloHCT) has increased substantially. Analyzing the long-term outcomes of these patients poses new methodological challenges since their population mortality (the risk of mortality that they would have faced in the absence of their disease and treatment) is non-negligible. Interpreting all non-relapse mortality (NRM) as treatment-related mortality cannot be justified in this older population. An additional problem in the analysis of the causes of death is that individual adjudication as to treatment-related complications is often impossible. Although more common in transplant recipients, both cardiac deaths and cancer also occur frequently in the general population in this age group. For these reasons, we propose to extend the analysis of standard outcomes with statistical methods to estimate the size of population- and treatment-related mortality in heterogeneous populations. We analyzed the outcomes of two large cohorts of patients whose data had been collected by the European Society for Blood and Marrow Transplantation: (1) myelodysplastic syndromes or secondary acute myeloid leukemia (n=6434, median age 56 years), and (2) chronic lymphocytic leukemia (n=2589, median age 55 years). We focused on patients who had survived for two years after alloHCT without relapse or progression. We used integrated methods from relative survival and multi-state models to estimate what proportion of mortality can be attributed to disease relapse and what to the expected general population mortality. A key assumption was that the risk of mortality of the patient cohorts, apart from that associated with the disease and its treatment, was similar to that of the age-, sex-, calendar year- and country-matched general population. The remaining mortality, excess non-relapse mortality, can then be interpreted as closely approximating to treatment-related mortality, where ‘treatment’ refers both to treatment prior to transplantation and to alloHCT and its sequelae. The figures illustrate the probabilities of these events for the MDS cohort. This non-parametric model was then extended to a Cox model where we also assessed the impact of age and other risk factors on the different components of mortality. Combining relative survival and multi-state models gives more insight into the different causes of mortality and the impact of age and other risk factors on them. We show that for older patients, a substantial part of NRM is actually attributable to population mortality, information which should hopefully lead to more informed discussions regarding the risk of alloHCT and improvement of long-term care. |
Full Text from ScienceDirect