We report a cost-effectiveness analysis comparing PBSCT and BMT in a cohort of pediatric patients. Cost benefit analyses measure both the costs and the effectiveness of alternative treatments in monetary units and thereby determine the net (social) benefit. Our study measures effectiveness in terms of clinical indicators/health outcomes and is the first to include such cost-effectiveness end point analyses out to 1 year of follow up. The clinical outcomes indicate that faster engraftment was observed in patients receiving PBSCT compared to recipients receiving BMT, but the frequency of infectious complications was higher in PBSCT recipients. Inpatient days and hospital costs were not statistically significantly different between the two transplant groups. At one year of follow up, patients in the BMT group had a higher treatment success rate and lower costs than the recipients of the PBSCT within the standard but not the high risk disease group, where the treatment success rate and the cumulative costs were lower in the PBSCT group compared to the BMT group.
Previous cost-effectiveness analyses of stem cell transplantation versus chemotherapy or no transplantation have largely studied adult populations with heterogeneous diseases [29
]. Several of these studies found SCT was cost-effective, as the intervention cost less than $50,000 per quality-adjusted life year (QALY) [32
]. Comparisons of the cost-effectiveness of PBSC versus and BM have been limited. Only one exclusively pediatric study analyzing cost-effectiveness of each stem cell source in children with hematological malignancy has been reported. This single institution study (Hospital Nino Jesus, Spain) of 25 patients accrued over 9 years found that overall costs at 100 days post transplant was lower for PBSCT than for BMT [36
]. Although more adult cost-effectiveness analyses have been published, only the acute costs were investigated [20
In a separate study, we analyzed the impact of stem cell source cells on disease-free survival in the pediatric transplant population at 1 year and 3 years of follow up, respectively. Our series showed there was a significant difference in the cumulative probabilities of disease-free survival between patients receiving BMT versus PBSCT by Kaplan-Meier method. However, multivariate Cox regression analysis showed no effect for the source of stem cells on treatment-related mortality, relapse or treatment failure overall whereas the HLA typing, disease status at transplant, the impact of CD34 selection and occurrence of acute or chronic GvHD were also included as independent variables. In the subgroups of patients with standard risk or high risk disease, the estimated probabilities of survival did not retain statistically significant differences between those receiving PBSCT compared to those receiving BMT. Of note, it was mostly stem cells from haploidentical donors that were CD34 selected, while non-selected PBSC were mainly used for matched unrelated donors. We subsequently performed multivariate analysis to adjust for independent effect of potential risk factors and found that CD34 selection was not associated with overall outcomes. Moreover, we analyzed the effects of age on the outcome observed with each stem cell source and found no significant correlation between graft type and age distribution. Ultimately, only the pre-transplant CMV seropositivity and differences in severity of underlying disease remained significant risk factors for disease-free survival.
By separating patients into those with standard and high risk disease, we found that despite the faster hematopoietic recovery observed in PBSCT recipients during initial transplant, there was no difference from BMT recipients in the time of hospitalization or costs incurred. In addition, during initial (short-term) follow up after hospital discharge to 100 days post transplant, the PBSCT group had higher costs and experienced a longer length of stay compared with BMT, although these differences did not reach statistical significance. Our results also showed a large difference in mean costs depending on the disease risk groups, a measure not available to the previous pediatric report [36
]. We analyzed the incremental cost-effectiveness ratio and showed that the treatment success rate was significantly different between disease risk groups. Although there was great variability in costs and the group sizes were relatively small, our results appeared to be robust in sensitivity analyses of the observed difference between each group.
We developed a cost prediction model from detailed accounting data to estimate costs for patients transplanted between January 2001 and September 2003. We also confirmed that during the study period there was no alteration in institutional allocation of expenses for the transplant unit. Hence, we have no reason to believe that any changes that may have occurred in hospital accounting practices would bias the cost comparisons between study groups. Based on the regression model we developed to impute cost data, the adjusted R2 were all well above .90 where less than 10% of variance would be explained by other additional variables, including those possibly outcome relevant variables. It was noted that the HLA-matched related donors (donor 2) have a lower negative coefficient compared to HLA-mismatched sibling donors (donor 1) as the latter have been perceived as being more costly. This internal inconsistency is likely due to the small numbers in this study, providing an idiosyncratic outcome that results from random effects in a small data set.
During the first year of follow up post transplant, patients mainly returned to the primary transplant clinic for standard of care. Inclusion of outpatient costs outside of the institution and homecare utilization would be expected to raise the total costs of follow up. One study examined the contribution of costs occurring between 3 to 6 months when patients returned to the care of their local physicians after 100 days, and found that they represented 3.8% of total costs [29
]. Although these data were not available at our institution, we would not expect the exclusion of those relatively minor cost estimates to result in significant differential increase in costs. Since the intent of our study was to analyze the total direct costs from the perspective of healthcare providers, we did not include the costs resulting from the time loss of patients or their families entered into the transplant program.
Although differences between institutions, and differing health care structures and pricing make direct comparisons with our study difficult, these data can be adjusted to factor in such distinctions. It is important to have precise estimates of hospital costs if the evaluation is to be accurate and so our study used micro-costing to measure all the direct medical costs of allogeneic transplantation for up to 1 year. These resource intensive transplant services may vary at regional and even local levels and our results may not be representative of other areas, even within the United States. Nonetheless, our detailed micro-costing and consistent cost categories allowed valid comparisons of resource requirements for the alternative therapies.
We investigated the total direct costs up to 1 year in all patients who received their primary transplant at our institution. Fourteen patients in our study had multiple transplants and were excluded from our primary analysis since these additional treatments are one of several salvage options for primary treatment failure and carry different, and non-comparable, expectations of cost and complexity irrespective of the stem cell source. For those patients who had multiple transplants after they relapsed from their initial transplant, the costs and outcomes information were subsequently measured in separate analyses. Since this was a retrospective observational study, accrual was subject to selection bias. However, treatment groups were similar in terms of their age, gender, ethnicity, and insurance coverage. We appreciate that fewer PBSCT than BMT recipients were studied, a difference that reflects clinical practice in pediatric SCT during the study period. It will certainly be helpful to reanalyze the data with a larger, more balanced cohort, since this would likely increase the confidence of our conclusions. Similarly, we followed the children for one year, an intermediate health care outcome, and a longer term study will be valuable to compare the longer term cost-effectiveness of each procedure.
The above limitations notwithstanding, cost-effectiveness analyses will likely become increasingly important as healthcare policies change. Our current incremental cost-effectiveness ratio and analysis of uncertainty suggest that allogeneic transplantation of bone marrow grafts was a more cost-effective treatment option compared to peripheral blood stem cells in patients with standard risk childhood acute leukemia disease. For high risk disease our data are less prescriptive, since the differences were more limited and the range of costs much larger. The comparative economic evaluation provides support for BMT for standard risk patients, but a great degree of uncertainty limits the clear advantage for either treatment option in patients with high risk disease. A larger and randomized controlled trial, especially in high risk patients, is essential to definitively demonstrate the long term cost-effectiveness of blood stem cells and bone marrow grafts for allogeneic transplantation of children.