To our knowledge, NC rates thus far have been reported for single clinical studies or large observational studies but not longitudinally or at the national level in the United States.5,6
NC rates are of special interest, as key clinical studies and guidelines have indicated that treatment patterns, particularly for lung cancer and breast cancer, have evolved to include increasingly myelosuppressive chemotherapy over the past 20 years. Current estimates of NC rates in the United States are thus likely underestimated as a result of the use of radiation therapy in combination with myelosuppressive chemotherapy in lung cancer. Treatment for NHL has been primarily based on cyclophosphamide, doxorubicin, vincristine, and prednisolone or cyclophosphamide, doxorubicin, vincristine, and prednisolone plus rituximab over this time span.7
During the study period, there was the important clinical development of dose-dense chemotherapy for breast cancer that incorporates the use of G-CSF prophylaxis.8
To examine these trends, we performed a descriptive cross-sectional analysis of NCs using NIS hospital discharge data from 1989 to 2007. Key findings showed that after excluding patients who received radiation therapy, on average, the estimated number of NCs in the total study population increased by approximately 30,000 cases per year. Furthermore, although the rate of discharges for the three study cancers combined remained relatively steady over time from 1989 to 2007, the rate of hospital discharges with NCs increased almost two-fold from 1989 to 1997. The use of G-CSF, especially after 2002, paralleled the clinical use of increasingly myelosuppressive chemotherapy regimens (dose-dense regimens, the use of taxanes, etc)9–11
and may well have mitigated the steeply rising annual incidence of hospitalizations from NCs from 1990 to 2000. However, the results of this analysis were derived from a combination of disparate data sources, which is a potential limitation of the assessment of the use of these agents.
The length of stay for hospital discharges with NCs decreased over time from 10.4 days in 1990 to 7.1 day in 2007. Of note, the length of stay for lung cancer may have been affected by changes in mortality rates. This change in length of stay may also be related to the advent of the diagnosis-related group payment system in the 1980s, and increased reimbursement restrictions put on the system since its introduction in 1983. In-hospital mortality rates with NCs for all study cancers fell from 10% in 1989 to 5.4% in 2007 at a fairly constant rate. Recent studies have suggested that G-CSF may influence mortality and may be one factor in reducing in-hospital mortality,12–14
although better inpatient treatment of patients with neutropenic infections could have contributed to this result.
There are several caveats that must be considered in interpreting these study findings. One fundamental source of bias is that this study relies on coded data and, hence, incorporates any coding errors that may have occurred. In a previous study, the sensitivity for defining neutropenia from ICD-9-CM coding was reported to be 80% when compared with other data sources.15
These data also reflect billing decisions, which may not always completely correlate with clinical assessments. In addition, the unit of analysis was the hospital discharge, not the patient; therefore, patient-level conclusions are out of scope for this study. Given the nature of the database, no adjustments could be made for cancer stage, treatment intent (curative v
palliative), or chemotherapy regimen. Also of note, the NC data were not adjusted for changes in chemotherapy use; shifts in care from one treatment modality to another (surgery, radiation, chemotherapy, or targeted therapy); or changes in population growth, cancer incidence, prevalence, or survival. These data are not easily accessible because the majority of NCs still require hospitalization,16,17
and large hospital discharge databases are often are not linkable to outpatient databases that would contain many of these treatment-related and patient-related data.
We sought to examine outpatient data using both the National Ambulatory Care Survey and National Hospital Ambulatory Care Survey databases to capture NCs observed in freestanding outpatient clinics and in clinics that are part of hospitals, respectively. However, the data obtained from these databases were not reliable because the number of cases was small and the SEs large. It may be that the study design did not permit sampling of an adequate number of oncologists, thus resulting in the number of NC cases being too low to report. Regardless, there was not a detectable trend in outpatient treatment of neutropenia in these data sets. This could be because most treatment of NCs occurs in hospitals.16,17
It is possible that other efforts to ameliorate NCs, such as dose-reduction strategies or other changes in chemotherapy regimens (especially in patients with metastatic disease or other factors predictive of poor outcome), may have affected the incidence of NCs.1,12,18–20
Our results indicate that the increased use of commonly prescribed myelosuppressive chemotherapies during this period is also of interest because increased absolute NC events would not be unexpected with some of these regimens.20
Although the analysis of the IMS Health Drug Distribution Database indicated increased use of most myelosuppressive chemotherapies, it is important to note that methotrexate is often prescribed for nononcology indications such as autoimmune diseases.21,22
To provide a broader context, we calculated 5-year cancer survival and prevalence rates from 1990 to 2001 based on data published in the Surveillance, Epidemiology and End Results Cancer Statistics Review, 1975 to 2006.23
These analyses showed that 5-year survival rates for all cancers, breast cancer, lung cancer, and NHL all increased from 1990 to 2001, with the smallest increase occurring with lung cancer and the greatest increase occurring with NHL. Specifically, survival for all cancers was up by 9.4% from 1990 to 2001, with the largest survival gains for NHL, which had a 16.6% increase, representing a 32.4% increase in survival. We likewise examined 5-year prevalence rates for all three study cancers over the time frame examined. The prevalence of lung cancer was unchanged from 1989 to 2006, whereas breast cancer increased by 18.2% and NHL increased by 50.0% compared with the 1989 rate. These data indicate that any decreases in NC hospital discharges over time were not due to a decrease in the number of patients with cancer.
In conclusion, our analyses demonstrated an increase in the number of hospitalizations for NCs from 1990 to 2000 with subsequent stabilization. This likely reflects the integrated effects of multiple factors considered by oncologists when making treatment decisions regarding neutropenia. These factors include the availability of increasingly myelosuppressive chemotherapy regimens, which may affect survival rates; widespread use of G-CSFs, such as pegfilgrastim and filgrastim, both prophylactically and therapeutically; and consideration of individual patient characteristics such as age and comorbid conditions. The relatively constant rate of NCs over the past several years indicates that it remains a significant clinical problem for patients. Additional research into how to optimize the balance of aggressive chemotherapy and prevention of febrile neutropenia may aid efforts to increase survival while minimizing the effects of febrile neutropenia.