­Resource Utilization and Cost Analysis By Treatment Arm on the Children’s Oncology Group AALL0232 Phase 3 High-Risk B-Precursor Acute Lymphoblastic Leukemia Trial: A Report from the Children’s Oncology Group

NCI-funded cooperative oncology group trials collect extensive biologic and outcome data but typically do not collect resource utilization (RU) or cost data. RU measures reflect therapy associated morbidity while cost data enable estimates of the financial burden of therapy. The ChildrenÕs Oncology Group (COG) trial AALL0232 used a 2x2 factorial randomization between 28 days of prednisone (PRED) or 14 days of dexamethasone (DEX) in Induction and high-dose (HD-MTX) or escalating ÒCapizziÓ methotrexate plus asparaginase (C-MTX) in Interim Maintenance I. Children ³10 years old were non-randomly assigned to PRED after May 2008 due to increased osteonecrosis on DEX. HD-MTX had superior event-free survival, and for patients <10 years on HD-MTX, DEX was superior. To estimate RU and inpatient costs on AALL0232, data from the Pediatric Health Information System (PHIS), a national database of daily inpatient billing data from 44 US hospitals, was merged with AALL0232 data. We hypothesized that inpatient RU and costs could be accurately estimated for AALL0232 in a merged COG-PHIS dataset and that RU and inpatient costs would be higher in children receiving HD-MTX, which requires hospitalization for administration.

COG provided an AALL0232 patient list that was probabilistically matched to PHIS patients by hospital, dates of diagnosis and birth, and sex. Patients were observed in PHIS data for 365 days post COG enrollment. Daily billing codes were used to tabulate RU and inpatient days. RU data include 30 variables from 4 major categories: laboratory, pharmacy, radiology, and procedures. Inpatient treatment costs were estimated by multiplying geographically adjusted charges by hospital-specific cost-to-charge ratios (RCC) and then adjusted to 2013 US dollars using the consumer price index.

Of 3083 patients enrolled on AALL0232, 935 (30.3%) were identified in PHIS. Two hospitals did not submit RCC to PHIS, so 20 patients were excluded from the cost analysis. As expected, age at enrollment differed significantly by study arm in the merged cohort (p<0.0001), while sex, race and ethnicity did not. Inpatient days were higher in the HD-MTX arms (median 53 days; interquartile range [IQR] 35-77) versus C-MTX (median 39 days; IQR 21-64.5; p<0.0001) and did not differ by steroid arm. Overall inpatient RU was similar in the HD-MTX and C-MTX arms, except for complete blood counts and antiemetics (Table 1). Children in the DEX arms had higher numbers of inpatient days of total, broad Gram-positive and antipseudomonal antibiotics than children on PRED (Table 2). There were no other RU differences that were both clinically and statistically significant, including intensive care RU. Inpatient costs were higher in the HD-MTX cohort and did not differ by steroid arm (Table 3).

These data demonstrate that inpatient RU and costs can be estimated in COG trials using external data sources and support the hypothesis that inpatient costs are higher in children receiving HD-MTX than C-MTX. However, the median cost difference between HD-MTX and C-MTX was modest ($31,286) and driven by length of hospital stay rather than increased RU. Children randomized to DEX had increased use of antibiotics, which may reflect an effect of DEX on infection risk. The merging of RU and cost data with COG clinical trial data enables treatment arm intensity and cost comparisons that provide important data on the medical and financial burdens of therapy. Work is ongoing to increase the percentage of trial patients with RU and cost data and to incorporate outpatient RU and cost data. Inclusion of RU and cost data in NCI-funded cooperative group trials should provide a more comprehensive assessment of randomized therapies and enable cost-effectiveness analyses.