Methods

All patients on the main protocol were eligible to be enrolled in the economic analysis. Prior to enrollment on the main protocol, all patients had to meet the protocol-defined eligibility criteria and sign the Institutional Review Board (IRB) approved informed consent documents that included consent to obtain hospital billing information for the economic component of the clinical trial.

The economic analysis encompassed two time segments: (1) the index hospital admission and (2) the 6-month follow-up period immediately following the index admission. Costs were measured from the hospital perspective and do not include physician professional fees or outpatient visits. All costs were adjusted to baseline year 2001 by means of the Hospital Producer Price Index published by the U.S. Bureau of Labor Statistics.⁶

Cost of index admission

Study sites were requested to obtain the Health Care Financing Administration (HCFA) Form UB-92 hospital summary discharge bill for each patient enrolled in the study. Billed charges were converted to costs using hospital-specific and year-specific Medicare cost-to-charge ratios.

Cost for follow-up hospital admission

Study sites were required to document hospital admissions during the 6-month follow-up period using a case report form (CRF) that recorded dates of admission and discharge, reason for admission and discharge status. However, collection of UB-92 billing forms was not requested because most of the admissions were expected to occur at hospitals not involved in the trial. Data from the CRF provided a basis for developing a file of similar admissions recorded in the Healthcare Cost and Utilization Project (HCUP) national hospital discharge database.⁷ Using the HCUP files for 1995–2000, statistical models were developed to estimate the cost of admission as a function of diagnoses (the patient's underlying form of leukemia and the specific reason for admission), patient age and gender, length of stay (LOS), discharge status and year of admission. Charge was converted to cost using the national average Medicare cost-to-charge ratio.

Analysis

LOS and cost of the index admission, the 6-month follow-up period and the combined index admission plus follow-up period were compared using unadjusted descriptive statistics. Tests of the hypothesis of no difference across treatment groups were performed using t-tests, ² tests, and the Wilcoxon signed rank test. Effect of T-cell depletion on LOS and cost of the index admission, and the combined index and follow-up period were evaluated using multivariate regression with log transformation of the dependent variable to adjust for right-skewed data.

Results

Clinical study findings

The T-Cell Depletion Study was conducted at 15 sites, with the first patient enrolled in March 1995 and the last patient enrolled in October 2000 and completing 6-month follow-up for the last evaluable patient on 30 April 2001. Characteristics of the study population are shown in Table 1. Of the 396 evaluable patients, 194 received TCD marrow and cyclosporine A (TCD), while 204 received M/C immunosuppression. Males comprised 55% of the study population. In all, 23% of patients were less than 18 years old, while 42% were more than 35 years old.

Table 1 - Characteristics of the study population.

Full table

Study centers differed substantially in terms of numbers of patients enrolled. Three large centers enrolled 56.1% (222/396) of study subjects, while nine centers enrolling 10 or more patients accounted for 93.9% (372/396) of enrollment. Treatment patterns across these nine centers were compared by examining the use of selected relatively costly resources such as hyperalimentation, intravenous antibiotic therapy and ventilator days during the index hospital admission (Table 2). While no formal statistical tests were performed, utilization of most of these resources, expressed as the number of units per 100 patient days, appeared comparable.

Table 2 - Comparison of average daily use of selected resources per 100 patient days, by number of patients enrolled at study site.

Full table

Economic study patient subset

Cost for the index admission was determined from the HCFA UB-92 summary discharge bill. Billing records were obtained for 62% (246/396) of patients eligible for analysis. The economic analysis as reported here is based on the subset of patients with billing records for the index admission (N=246). In a few cases, the index admission actually consisted of two or three sequential admissions separated by periods of a few days; LOS and charges were aggregated to create a single index admission bill for each patient. Comparison of characteristics for patients with and without index admission billing records is shown in Table 3. No statistically significant differences were observed in age, gender, proportion of patients with diagnosis of CML, or treatment arm. However, index admission LOS was somewhat longer for the group without billing data (median 42 vs 36 days, P<0.01).

Table 3 - Comparison of subjects with and without billing record (UB92) for index admission.

Full table

Figure 1 illustrates the relative magnitude of components of treatment costs for the full study cohort based on the billed charges by hospital revenue center as reported on the UB-92 summary bills. Charges for room and board, including admissions to intensive care units, and pharmacy charges accounted for nearly two-thirds of total charges as observed in other economic analyses of stem cell transplantation published in Kline et al's study (1998).^{2, 8}

Figure 1.

Relative value of resources used during the index admission.

Full figure and legend (68K)

LOS and cost of the index hospital admission

LOS and cost of the index admission for the subset of patients with billing data are shown in Table 4. Both LOS and cost of admission appeared somewhat lower for the TCD group relative to the M/C group. However, LOS was not statistically different (P=0.55) with median 34.0 days for the TCD group vs 39.0 in the M/C group. Similarly, no statistical difference (P=0.18) was observed in the median cost for the index admission: $102 583 for the TCD group vs $115 413 in the M/C group.

Table 4 - Comparison of length of stay and cost of hospital admission by study treatment group.

Full table

LOS and cost of follow-up hospital admission

Patients were actively followed during the 6 months subsequent to discharge from the index hospital admission. Figures for LOS and cost of admission for the six-month follow-up period (Table 4) are based on the economic study subset of patients with UB-92 s who survived the index admission: 62.2% (74/119) of the TCD group and 57.5% (73/127) of the M/C group. These data may reflect multiple admissions since patients in the TCD group were hospitalized an average of 3.7 times, while the conventional patients were admitted 3.5 times. LOS was not statistically different (P=0.31), with medians of 17.5 and 16.0 days for the TCD and M/C groups, respectively. A cost for each follow-up admission was imputed as previously described. Median follow-up hospital cost was $46 143 in the TCD group and $26 888 for the M/C group (P=0.36).

LOS and cost of admission for the combined index and follow-up periods

Combined LOS and costs for hospital admissions during the index treatment and 6-month follow-up period are also shown in Table 4. Median LOS (eg, total hospital days) was similar across groups, 47.0 vs 52.0 days (P=0.72) for the TCD and M/C groups, respectively. The total cost for hospital admissions was also comparable: $145 115 for the TCD group vs $141 981 for the M/C group (P=0.63).

Effect of TCD on hospital LOS and costs

Determinants of LOS and cost for the index admission and the combined index admission and follow-up period admissions were explored using multivariate regression for the subset of patients with index admission billing records. The regression models controlled for treatment group assignment, patient age, race and gender, clinical features, and transplant center. These results are shown in Table 5. Regression parameters have been transformed into the percentage change in either LOS or cost as a function of the specified categorical variable.

Table 5 - Determinants of length of stay and cost for the index admission and for the combined index admission plus 6 month follow-up period.

Full table

TCD was associated with a 12.1% reduction in LOS for the index admission (95% CI -19.4%, -4.3%), with a 5.3% reduction in cost of the index admission (95% CI -15.8%, 6.4%). When considering the combined index admission and follow-up periods, TCD was associated with a negligible 3.2% (95% CI -13.4%, +8.3%) decrease in total hospital days and 2.8% decrease in cost. Independent of treatment assignment, HLA matching was associated with an 8.6% (95% CI -17.4%, +1.2%) reduction in the index admission LOS, while cost was decreased by 15.8% (95% CI -26.7%, -3.3%). Over the combined index and follow-up period, LOS for patients with a 6 of 6 HLA match was decreased by only 2.2% (95% CI -10.0, +17.4%), while cost declined by 10.0% (95% CI -23.1, +5.3%).

Survival vs cost

Cost of treatment could be associated with duration of survival, such that patients surviving longer incur higher costs. In that case, a relative survival advantage for one of the study treatment groups could result in higher costs. To explore the relationship between cost and survival, the mean cost of treatment was calculated for subgroups categorized by duration of survival (Figure 2). Costs for the two treatment groups were quite similar within survival duration categories. However, the relationship of cost vs survival appeared to follow a 'U-shape' pattern, where average costs increased with duration of survival up to the 121–150 day period and then began to decrease. This finding was substantiated in a linear regression that modeled cost as a parabolic (quadratic) function of survival duration (not shown).

Figure 2.

Total cost for hospital admissions by duration of survival (days) following randomization. Bars denote the interquartile range bounded by the 25th and 75th percentiles of total costs among patients in the survival duration category. Figures above bars indicate the number of patients within each group.

Full figure and legend (21K)

Impact of GVHD on hospital costs

Patients in the TCD group were less likely to contract severe cases of acute GVHD than those in the M/C group. By day 100 post-transplant, 64.8% of the TCD patients had experienced grades 0-I acute GVHD compared with 41.4% in the M/C group. By contrast, 30.4% of patients in the M/C group experienced grades III-IV acute GHVD compared with 16.2% among TCD patients. Cost of the 6-month treatment episode by highest severity level of GVHD is depicted in Figure 3 for each of the two treatment groups. Average costs rose progressively with increasing level of GVHD severity. However, costs within severity levels were similar across treatment groups. Severe levels of acute GVHD added a substantial amount to the cost of treatment. Relative to a patient who did not experience acute GVHD, cost in the first 6 months post-transplant, adjusted for treatment group and HLA matching, was increased by $60 419 for grade III acute GVHD and by $76 070 for grade IV acute GVHD (P<0.05).

Figure 3.

Total cost for the combined index admission and 6-month followup period by highest severity level of GVHD. Bars denote the interquartile range bounded by the 25th and 75th percentiles of total costs among patients in the GVHD category. Figures above bars indicate the number of patients within each group.

Full figure and legend (19K)

Cost-effectiveness

Cost-effectiveness analysis measures the incremental cost and benefit of an intervention relative to a comparator. Cost-effectiveness for TCD relative to unmodified donor marrow in the TCD Study is calculated as:

The preferred metric for benefit is the quality-adjusted life-year (QALY), which reflects the period of survival weighted by patient preference for the health states experienced over that period.⁹ The TCD Study included assessment of health-related quality of life at baseline and over the course of follow-up. However, no differences in quality of life that could be extrapolated to preference weights were observed by 1 year post-transplant.¹⁰

An alternative measure of benefit is incremental survival. No difference in 3-year survival was recorded in the TCD Study. In situations where there is no difference in costs and no difference in benefits, cost-effectiveness analysis is not informative.

Discussion

The TCD trial is the largest multi-center, randomized, prospective trial in URD marrow transplant, with 3-year DFS as the primary end point performed to date. The trial included an economic analysis component based on standard methods for cost-effectiveness analysis in transplantation for leukemias and lymphoma.¹¹ While other studies have estimated the costs of URD transplantation,^{12, 13} there has only been one prior analysis of the impact of TCD on costs and outcomes in URD recipients.¹ In that study, however, patients were not randomized, but were selected for either TCD or standard immunosuppressive therapy based on the attending physician recommendation and patient preferences after discussion of the risks and benefits of each method.²

In concert with the study by Lee et al, we found a trend to lower cost of the index hospital admission for patients randomized to TCD relative to patients randomized to receive an unmodified transplant. Since room and pharmacy costs comprised nearly two-thirds of the total costs in our trial, the higher cost of the index hospitalization, with the longer LOS for patients receiving unmodified transplants, can be attributed almost entirely to the room and pharmacy costs.

As demonstrated in this study, there was no difference in costs between patients undergoing TCD vs unmodified transplants within each grade of acute GVHD. However, the severity of acute GVHD, as it occurs, adds significantly to the costs of transplant. Since there were fewer patients in the TCD arm who developed grades III and IV acute GVHD, there were less total costs associated with the initial hospital stay in the TCD arm. However, during the first 6 months following transplantation, there was a statistically insignificant increase in number of days of hospitalization relative to the unmodified group, which added to the costs of recipients of TCD transplants. This translated into an absence of overall difference in LOS or costs associated with TCD vs unmodified transplants during the combined index hospital admission and follow-up period. Savings on reduced cost for treating acute GVHD were likely offset by an increase in serious infections on the TCD arm.¹⁴

When analyzed by duration of survival, there were no statistical differences between the TCD and unmodified arms. For recipients who lived less than 60 days, there was no difference in cost by treatment arm. Similarly, for patients who lived longer but died, the costs of the overall care were increased, but there were no statistical differences between the TCD and unmodified arms at any time period. However, surviving recipients of an unmodified transplant tended to have higher costs associated with that survival.

A trend to lower costs was observed when donors and recipients were more closely HLA matched (6/6 vs 5/6). During the course of this study, matching criteria required serologic testing at HLA-A, -B loci and high-resolution molecular match for DRB1. The effect on cost of transplantation based on the incidence of acute and chronic GVHD, with the use of more sensitive molecular based HLA typing, and requirements for matching at HLA-C, and -DQ, to obtain more closely matched donor–recipient pairs, remains to be determined.

None of the costs associated with preparing the donor marrow for infusion in the TCD arm were accounted for in this study. Two methods of TCD (elutriation with CD34 add back and T10B9 monoclonal antibody plus complement) were used during the conduct of this trial.³ While TCD did decrease the incidence of significant acute GVHD, the mean difference in cost for those patients was not different between arms.

Treatment outcomes may be associated with transplant center volume. Previous studies of therapy such as management of pancreatic cancer have reported improved patient outcomes and lower costs for hospitals performing high volumes of a particular procedure relative to low-volume providers.¹⁵ To explore the cost–volume relationship in the TCD trial, we expanded the multivariate regression as reported in Table 5 to explore the relationship between index hospital admission and total number of study patients enrolled. Results suggested that enrolling larger numbers of patients was associated with lower cost of the index admission. Note that this finding is only suggestive, since numbers of patients enrolled in the trial may not reflect the extent of institutional experience in bone marrow transplantation.

In summary, while TCD of the donor graft may reduce short-term costs associated with unrelated bone marrow transplantation, long-term complications leading to frequent hospitalizations and higher average number of hospital days in the first 6 months following transplantation appear to offset any savings seen with the initial transplant stay. More sensitive HLA typing methods for HLA-A and HLA-B, as well as matching at HLA-C and HLA-DQ to obtain more closely matched donor–recipient pairs, may be an alternative strategy to lowering costs with either TCD or unmodified transplants for individuals who may benefit from an unrelated hematopoietic stem cell transplant.

References

References

Acknowledgements

We thank the patients who willingly entered this large clinical trial, the physicians, nurses, and other support staff who cared for them during the transplant procedure, and the clinical investigation team at each participating institution for data collection and follow-up. We also thank the National Heart, Lung, and Blood Institute (NHLBI) for support of this trial. In addition to the authors, the following transplant centers, study physicians, and experts contributed to this study: University of Minnesota (N=103; John E Wagner, MD and Stella M Davies, MBBS, PhD), Memorial Sloan-Kettering Cancer Center (N=70; Esperanza Papadopoulos, MD, Richard O'Reilly, MD), Medical College of Virginia (N=53), Wake-Forest University-School of Medicine (N=36), University of Nebraska (N=34; Stephen Pavletic, MD, Michael Bishop, MD), University of Utah (N=33; Finn Bo Petersen, MD), Stanford University (N=25; Robert Negrin, MD), University of Iowa (N=19; Robert Gingrich, MD), University of South Carolina (N=13), Ohio State University (N=6; Edward Copelan, MD), Duke University (N=6; Joanne Kurtzberg, MD), University of Kentucky (N=5; John S Thompson, MD, Gordon Phillips, MD), Medical College of Wisconsin (N=4; James Casper, MD, Carolyn Keever-Taylor, MD, William Drobyski, MD, Neal Flomenberg, MD), Western Pennsylvania Hospital (N=2; Richard Shattuck, MD), and University of Pittsburgh (N=1; Albert Donnenberg, PhD), the EMMES Corporation (Donald Stablein, PhD, Adam Mendizabal, MS, Elizabeth Wagner, MPH), NHLBI (LeeAnn Jensen, PhD, Nancy Geller, PhD, Paul McCurdy, MD), MEDTAP International (Sandra Macker). This trial was supported by a contract from the National Heart, Lung and Blood Institute (N01-HB-47094 (GL, SLC and ME), N01-HB-47097 (DH, PB, JHD, and SY), N01-HB-47098 (NAK), and N01-HB-47095 (DW)).