Cancer therapy has experienced great changes in the last 20 years, especially with the use of autologous bone marrow transplantation (ABMT) following ablative chemotherapy. From a clinical perspective, treatment outcomes are greatly improved. Unfortunately, with the introduction of new technologies, health care costs have skyrocketed. The total transplant treatment course has resulted in extremely high health resource utilization, as measured by a plethora of variables, from cost of hospital days to drug interventions, to the technology itself. Health care economists and researchers ask, What is the added value of improved technology and pharmacopeia? Is the cure worth the cost?
A significant factor contributing to this cost model (and poor clinical outcome) has been the secondary complications of transplantation resulting from severe neutropenia due to ablative chemotherapy. A significant recent discovery that may break the cost trajectory is the introduction of a human granulocyte colony-stimulating factor (G-CSF) into the armamentarium of cancer treatment therapies. Filgrastim(Drug information on filgrastim) (Neupogen) is designed to "stimulate" the recovery of leukocytes following aggressive chemotherapy. Clinicians have quickly seen the potential of G-CSF for decreasing the post-ABMT recovery period and modifying the problematic morbidity and secondary complications associated with neutropenia.
Since February, 1991, G-CSF has been used in FDA-approved indications in conjunction with ABMT. While clinicians may debate the potential impact of G-CSF on overall health care costs for cancer therapy, as well as on improved outcomes, to date, no systematic study has been undertaken.
In this study, our specific aim is to examine the effect of G-CSF on the use of health care resources during the inpatient phase of treatment and recovery post-transplant. This period has been chosen since it is generally acknowledged as the most resource-intensive phase. The study groups are stage IIIa, IIIb, and IV breast cancer patients. Our specific research question is whether the introduction of G-CSF into the protocol following ablative chemotherapy with ABMT significantly decreases the consumption of health care resources and subsequent care.
The introduction of G-CSF into the treatment course can potentially reduce the utilization of health care resources in the inpatient setting by reducing the length and intensity of the high-resource utilization periods. Further, earlier absolute neutrophil count (ANC) recovery may open up the opportunity for effective and efficient treatment in ambulatory settings.
The specific study hypothesis is as follows: The use of G-CSF in conjunction with ABMT will significantly decrease the number of patient days of severe immunocompromise, the number of days of triple antibiotic therapy, the number of days in which the ANC is less than 500/mm³, and the overall length of stay.
For this pilot study, we used a retrospective design. All patients treated for breast cancer during the last 6 years at one center (see Table 1) were studied with respect to measures related to inpatient length of stay. As a major insurance provider for our region, we were able to arrange access to the medical records of all the breast cancer patients treated since 1989 by a single major provider of oncology services. To maintain the homogeneity of the study group, we chose to analyze the patients of one physician group, which is the major oncology practice group in this region. We assume that the disease of breast cancer is similar across all the cases.
For each decision regarding this study, we consulted with a cross-section of expert oncologists to determine the relevant approach to our research question. Our first decision was to study only solid tumor malignancies, focusing exclusively on stage IIIa, IIIb, and IV breast cancer patients who underwent autologous transplantation between 1989 and 1993. These years comprise the historical period immediately prior to (1989-1991) and following (1991-1993) FDA approval of G-CSF. Mirroring the historical changes in bone marrow therapeutics since 1989, our treatment groups consist of various combinations of purged/nonpurged bone marrow and peripheral stem cell reinfusion, both with and without treatment with G-CSF (Table 1).
In defining the major outcome variables, the investigators first considered using length of stay in the acute care facility. However, the number of days a patient spends in treatment may not be as significant in determining cost of care as the number of days spent in a high resource utilization environment, such as an intensive care unit. The investigators mapped the admission course of subjects into three major areas: baseline, days of high resource utilization, and recovery period. Focusing on the high resource period of an admission, we asked, on what aspects of the treatment protocol would the earlier recovery of leukocytes, using G-CSF, have the most impact?
We have chosen outcome variables that, in the view of practicing oncologists and consultants, would be most affected by a decreased ANC recovery period. Besides total length of stay, our outcome variables (Table 2) specifically apply to the high resource utilization periods of severe neutropenia (ANC less than 500/mm³) and triple antibiotic therapy.
Each of these outcome variables is a measure of severity of condition and the corresponding need for intensive medical support and management. For example, sustained neutropenia generally determines the number of days in isolation and the number of days a patient is on triple antibiotic therapy, each of which are cost-intensive therapeutic measures.
Frequency distributions, means, medians, and standard deviations were calculated for each treatment group. Statistical tests consisted of one-way analysis of variance followed by Fisher's LSD method of multiple comparisons. Box plots were used to explore whether groups had similar spreads and whether any outliers were present in the data. Variances between groups were tested using Bartlett's test, and residual diagnostic techniques were employed to determine if the assumptions of analysis of variance were valid. If any assumptions were invalid and could not be corrected using transformations, the Kruskal-Wallis one-way nonparametric test was used to test differences between groups. This test was followed by the Mann-Whitney U test for individual comparisons. P values at this point were considered significant if they were less than 0.025. Statistical analysis was performed using SAS Institute (Cary, NC) statistical software.
Using these techniques, we tested the hypothesis that the use of G-CSF significantly decreases the utilization of health care resources, operationalized in this study as the outcome variables for our sample of breast cancer patients from one provider site.
As seen in Table 3, total length of stay showed a significant difference across all groups (P = .003) toward a shorter admission, from a mean of 43 days for group I to 24.9 days for group V. Further, group III (purged bone marrow cells supported with G-CSF) showed a 20% reduction in length of stay compared with group I (purged marrow with no G-CSF) (P = .05). The impact of fewer cells available for engraftment, due to the toxic effects of the purging process, is apparently ameliorated by G-CSF.
Group IV (nonpurged with G-CSF) showed a 17% length of stay reduction (mean stay, 32.8 days), compared with group II (nonpurged, no G-CSF; mean stay, 39.4 days), although this difference was not significant. Group V had the shortest length of stay (mean, 24.9 days), which reflects the advance in the recovery capability of peripheral stem cells.
Number of days with ANC count below 500 showed a significant difference across all four bone marrow comparison groups toward fewer days with neutropenia (P = .001) with G-CSF. In terms of significant individual group comparisons, group IV (nonpurged with G-CSF) had a 36% reduction in number of days with ANC less than 500, compared with group III (purged with G-CSF) (P = .004). Groups I and III (P = .02), and groups II and IV (P = .009) comparisons were also significant.
Despite possible variation in physician practice styles and individual patient characteristics, it generally follows that sustained neutrophil suppression (ANC below 500) most closely correlates with severity of condition and the use of health care resources. The reduction in days of sustained neutropenia, and corresponding reduction in high resource utilization days, is convincing evidence of the impact of G-CSF on health care resource utilization, at least during episodes immediately following ABMT.
Analysis of total days on G-CSF was limited to the comparison among groups III, IV, and V. As Table 3 shows, there was no significant difference in total days on G-CSF between groups III and IV, and also between groups IV and V. However, the comparisons between groups III and V showed a significant decrease in number of days on G-CSF (P = .001). This result suggests a decreasing trend in length of stay for the peripheral stem cell treatment group, which, in turn, suggests a more rapid recovery for patients transplanted with peripheral stem cells and a need to customize dosing duration of G-CSF in these patients.
Total cumulative unit days of antibiotic use and total cumulative unit days of antibiotic use divided by 3 were significantly different across all groups (P = .005). Groups I and III showed a significant difference, with mean values of 72.4 days and 50.3 days, or a 31% reduction in days on triple antibiotics for the purged cell group with G-CSF use. In the nonpurged groups (groups II and IV), use of G-CSF led to a 19% decline in days on triple antibiotics, but these results were not statistically significant, due to the variability within these treatment groups. This result is consistent with a greater impact of G-CSF on the recovery time of neutrophils in the purged cell groups (P = .05) than in the nonpurged treatment groups.
Group V (peripheral stem cell with G-CSF) had a mean of 24.5 days on triple antibiotics, the most dramatic decrease in resource consumption among any of the groups, and this reflects the enhanced rate of recovery of patients in this category. As expected, total days of antibiotics divided by 3 showed the same results as total cumulative unit days of antibiotic use.
The use of G-CSF in supporting recovery for patients undergoing high-dose chemotherapy/autologous transplantation for stage III and IV breast cancer produces several key advantages. In comparing both purged and nonpurged groups, we showed an improvement in reducing days of ANC below 500 with use of G-CSF.
Comparisons between groups I and III showed an improvement in both length of stay and days on triple antibiotics with G-CSF use. While we were unable to assign a statistically significant result to the other outcome variables, we saw a significant and consistent trend in favor of the use of G-CSF between groups II and IV in reducing length of stay and days of triple antibiotics. For groups III, IV, and V, days on growth factor fell significantly across the groups (from a mean of 27.2 for group III, to 23.3 for group IV, to 16.5 for group V.
Any limitations of the study may be attributed to the natural variance within the treatment groups of breast cancer patients. A larger sample size may help resolve these limitations in future study designs but does not appear warranted, given the extremely low P values found.
Since the use of G-CSF with peripheral stem cell reinfusion has become the standard of care with autologous transplant patients, we will not be able to assess the resource utilization impact of stem cell transplantation without G-CSF.
The reduction in days of treatment for severe neutropenia may reduce direct costs of care and depress the secondary costs often associated with high levels of morbidity and mortality (which were greater than 10% and as high as 30% during the initial [pre-G-CSF] ABMT trials in 1990 and 1991). These questions are ripe for direct study. At our study center, ABMT costs have dropped by more than 50% since 1990, perhaps as a result of pharmaceutical intervention and the use of peripheral stem cell transplantation.
In terms of the clinical impact of our results, the dependable and fast ANC recovery may change the conservative nature of cancer chemotherapy. If clinicians have the means to stimulate a faster ANC recovery, we can potentially increase the intensity and scope of aggressive chemotherapy without irrevocably compromising the patient's status and, secondarily, increasing overall costs.
There are numerous further implications of a faster ANC recovery. The reduction in morbidity and mortality using G-CSF may suggest the greater application of ABMT therapy to patients in earlier stages of disease and to patients seeking palliative care. In addition, it may be possible for a greater proportion of care days to take place in the less costly ambulatory setting. Finally, it may allow the use of a less conservative protocol with greater chemotherapeutic doses at earlier stages in the illness, which could potentially improve the disease trajectory and outcome.
The increasing use of peripheral stem cells (group V) has propelled the progress and value of autologous transplantation to the next generation. The use of these peripheral stem cells, in conjunction with current and developing cytokines, will allow for far greater kill ratios among cancer cells, while avoiding resistance and maintaining a minimal impact of treatment-related morbidity and mortality. To fully examine the impact of G-CSF on the treatment of breast cancer patients post-ABMT, an appropriate next step would be a study of the impact of peripheral stem cell use without G-CSF. However, as mentioned above, given the current standard of care, this approach would not be medically acceptable.
Future studies extending our original hypotheses are possible. Nevertheless, our data clearly indicate the value of G-CSF in reducing the use of costly acute care resources. G-CSF plays a valuable role in both the clinical and utilization management of patients undergoing autologous transplantation. As health care reform and integrated treatment systems evolve, the cost efficiency of pharmaceuticals in reducing the use of high-intensity health care resources should be extremely useful information, directing the decision making of physicians and other health care leaders.
The authors wish to point out that every effort was made to minimize the effects of nonaligned variables within our study. It must be recognized that during the period of our investigation advances in both clinical care and supporting technology may have contributed to the reduction in length of stay and other measures along with the use of G-CSF. However, the authors continue to believe that the use of G-CSF was the major contributor to any meaningful reductions in health resource utilization.