|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2000 American Society for Clinical Oncology Growth Factor Usage Patterns and Outcomes in the Community Setting: Collection Through a Practice-Based Computerized Clinical Information SystemByFrom OnCare IncAtlanta, GA. Address reprint requests to Grant Swanson, MD, 261 El Dorado, Suite 202, Monterey, CA 93940.
PURPOSE: Although use of colony-stimulating factor (CSF) is widespread and guidelines for use have been disseminated, actual practice patterns of medical oncologists are unknown. The purpose of this study was to collect these data using an office-based computerized clinical information system. PATIENTS AND METHODS: Data were collected on patients at 10 community-based oncology practices. Information regarding CSF use was captured at the time of prescribing through a computerized clinical support tool and stored in a data warehouse, and an analysis was carried out retrospectively. RESULTS: A total of 6,813 cancer regimens administered to 5,034 patients were evaluated for growth factor use. Overall, CSFs were used in 14% of regimens, with breast, lymphoma, lung, and ovarian being the most common cancers for which CSFs were used. In 49.4% of regimens, CSF was initiated during cycle 1, with an average duration of 1 week, and was used in two or three cycles per regimen. Afebrile neutropenia is rarely followed by CSF initiation. Granulocyte colony-stimulating factor (G-CSF) is associated with fewer dose adjustments, delays, and hospitalizations when compared with granulocyte-macrophage colony stimulating factor (GM-CSF). There is wide variation among oncologists in CSF use, and several substantial differences were noted between the prescribing behavior of American Society of Clinical Oncology (ASCO) survey–reported oncologists and actual clinical practice, as captured by the computerized clinical information system. CONCLUSION: Computerized clinical information systems can collect detailed information regarding practice patterns of medical oncologists. ASCO physician practice survey data do not accurately reflect actual practice patterns and must be interpreted with caution. Substantial deviations from ASCO growth factor guidelines remain, and oncologists’ use of CSFs demonstrates wide variation. There may be important clinical differences between G-CSF and GM-CSF, but definitive phase III trials are needed for confirmation.
BECAUSE OF THE clinical importance, widespread use, and significant costs of colony-stimulating factors (CSFs), one of the first clinical guidelines prepared by the American Society of Clinical Oncology (ASCO) dealt with CSF use.1,2 Two surveys have been conducted and the results published to examine purported use of these agents by oncologists in clinical practice, one before3 and one subsequent to4 guideline publication. To assess practice patterns, the physicians surveyed were asked to respond to a series of clinical vignettes, and responses were compared with guideline recommendations. Comparison of these two surveys would imply that after guideline dissemination, there was a shift in practice patterns toward greater compliance with ASCO recommendations regarding CSF use, with less use occurring in the setting of afebrile neutropenia and decreased secondary prophylaxis after afebrile neutropenia. The obvious shortcoming of this analysis is the uncertainty over the extent to which these responses can be correlated with actual clinical practice. Collection of this data is hampered by the absence of sophisticated information systems that are capable of extracting detailed clinical information in most practice settings. Without these systems in place, the practice patterns of oncologists outside the context of clinical trials remains largely unknown. OnCare (Atlanta, GA) is an oncology physician practice management group with a central focus on the integration of information technology with clinical practice. This report represents the actual clinical use of CSFs by oncologists in the OnCare network as documented in the OPUS Matrix System (Axion Healthcare, Woodside, CA), a point-of-care clinical decision support tool that provides computerized treatment guidelines and clinical drug information to the medical oncology practice while collecting patient care data.
Patient Selection This study is a retrospective analysis of data that were captured electronically at the time of office-based outpatient encounters. Participating physicians are members of OnCare, a national physician practice management group consisting exclusively of adult medical, radiation, and gynecologic oncologists. All patients seen during the study period (January 1996 through March 1998) were eligible for analysis. All patients who received CSF, excluding patients with leukemia or myelodysplasia, were included in this analysis.
Computerized Clinical Information System
Statistical Methods
Patient Characteristics We evaluated 6,813 cancer regimens (defined as a specific drug or drug combination administered on a scheduled repetitive basis) that were administered to 5,034 patients and found that growth factors were used in 46 different cancer diagnoses. The diagnoses for which 10 or more regimens used CSFs are listed in Table 1. Overall, of 6,813 regimens that were evaluated in this study, growth factors were administered in 950 (14%). Table 2 lists those diagnoses with the highest percentage of growth factor utilization.
Regimen Summary The most common chemotherapy regimens that incorporated CSF use are listed in Table 3. These 11 regimens account for 52% of all regimens with which growth factors were used. Table 4 shows that the average number of CSF doses per chemotherapy cycle (defined as one sequence in a series of repetitively administered drugs) is 6.2, with CSF use occurring in two to three cycles per regimen (15.3 doses/regimen).
Administration Timing Table 5 lists the number of the cycle in which CSF was initiated, with one half of all patients receiving their first CSF in cycle 1. Fig 1 shows the wide distribution of days on which CSF was initiated, with peaks immediately after chemotherapy and at day 7. For patients who had CSF initiated within 6 days of chemotherapy, the average starting time was 1.94 days after chemotherapy. For those beyond 6 days, it was 16.8 days.
Effect of Prior Chemotherapy on CSF Use To evaluate whether there is a tendency toward increased CSF use with increased number of prior chemotherapy regimens, we examined four common diagnoses. As listed in Table 6, there was no obvious trend toward either increased frequency of use or increased number of doses/cycle when comparing initial therapy with first or second salvage regimens.
Physician Response to Low Absolute Neutrophil Count Without Fever We looked at patients who developed neutropenia without fever and assessed physicians’ use of CSF in the next chemotherapy cycle. Patients were divided into those who received CSF at the time of neutropenia and those who were not receiving CSF at the time of neutropenia. In those patients who were already receiving CSF at the time of afebrile neutropenia and who went on to receive another cycle of chemotherapy (Table 7), 70% continued receiving CSF on the next cycle. The remainder either had the regimen ended, the next cycle deleted, or CSF stopped. In patients who were not already receiving CSFs who developed afebrile neutropenia at nadir and who went on to receive another cycle of chemotherapy (Table 8), only 9% had CSF started on the next cycle, with fully 91% of patients not receiving CSF on the next cycle. There was also a small number of patients who had the regimen ended or the cycle deleted.
G-CSF Versus GM-CSF Table 9 lists all patients who received growth factors for whom a delay in therapy or dose adjustment occurred secondary to neutropenia; this frequency is compared in regimens with G-CSF versus GM-CSF. Also included are hospitalizations for febrile neutropenia. G-CSF resulted in fewer dose adjustments and delays for both afebrile neutropenia and febrile neutropenia than did GM-CSF. There were more hospitalized patients who were being treated with GM-CSF than with G-CSF for febrile neutropenia. All comparisons were statistically significant. The odds ratios express that patients who receive GM-CSF are 6.25 times more likely to be hospitalized than patients on G-CSF, 3.85 times more likely to have a dose delay associated with febrile neutropenia, and 1.79 times more likely to have a dose delay associated with afebrile neutropenia. Patients who receive GM-CSF also are 1.67 times more likely to have a dose adjustment associated with afebrile neutropenia and 7.69 times more likely to have a dose delay associated with febrile neutropenia than patients who receive G-CSF.
Practice Variation We compared the frequency of CSF use among 10 different oncology practices and found that incorporation of CSF varied from 4% to 27% of all patient regimens administered (Table 10). CSF use, when evaluated at the level of the individual oncologist within a practice, also varied greatly. For example, within a single practice, CSF use by individual physician ranged from no use by one physician to use with 44% of chemotherapy regimens by another physician (Table 10).
Comparison With ASCO Survey We compared the ASCO survey–reported data with actual oncology practice behavior in our study using comparable categories. Every ASCO scenario could not be included, because our data represent only outpatient encounters and do not include physician decisions in the inpatient setting. Oncologists in this study used a higher percentage of GM-CSF than was reported in either survey (Table 11). Primary prophylaxis in the adjuvant treatment of breast cancer showed very similar clinical use to that reported in the surveys. However, in clinical practice, use of growth factors as primary prophylaxis in salvage treatment of ovarian cancer was far less than would be anticipated from the surveys (16% v 53%/39% in the two surveys). In addition, initiation of CSFs as secondary prophylaxis in the cycle after afebrile neutropenia was substantially less in this study than that which was reported in both surveys (7% v 44%/36% in the two surveys).
The clinical and economic impacts of growth factors have been the subject of intense interest since the introduction of these products in 1991.1,2,5-9 Although explicit evidence-based guidelines have been published regarding the use of CSFs by ASCO,1,2 recent reviews have argued that there is little demonstrated benefit to the use of these agents in many clinical practice settings.10 Actual usage patterns of CSFs by medical oncologists outside the context of a clinical trial are unknown. Two surveys conducted by ASCO have attempted to measure physician use of these agents as well as the impact of the publication of guideline recommendations.3,4 By completing multiple-choice questions related to clinical scenarios, data have been presented that may indicate increased compliance with the guidelines from 1994 (just before guideline publication) to 1997 (after a 1996 update). The initial survey3 found that although the majority of oncologists reported use of CSFs that was consistent with the ASCO guidelines, over one half of oncologists used these agents in the setting of afebrile neutropenia, and, in certain clinical scenarios, a substantial percentage of oncologists would initiate CSF support as secondary prophylaxis after afebrile neutropenia. The follow-up survey found appropriate decreases in preferences for use in both of these clinical settings.4 Although these data are provocative, the concordance of actual oncologic practice with responses to brief idealized clinical vignettes is unknown. Indeed, it was suggested that, "Data on actual practice patterns are needed because what physicians say they do and what they actually do may be different."4 This study begins to fill that gap. All data on CSF use were entered into the electronic system at the time of clinical decision making. Hence, we were not subjected to the problems associated with retrospective reviews of paper charts with incomplete and illegible data entry. There are several important conclusions from this study. CSFs are being used in the treatment of the vast majority of solid malignancies. In this study, they were incorporated into 14% of all cancer regimens, with breast, non-Hodgkin’s lymphoma, lung, and ovarian cancers representing the most common diagnoses with which CSFs were used. Likewise, CSFs have been incorporated into a large number of chemotherapy regimens, and the average duration of use is approximately 1 week. The distribution of time of CSF initiation likely represents the clinical scenarios in which these agents are used. The early peak of 1.94 days after chemotherapy is correlated with conventional CSF use immediately after completion of chemotherapy. The population of patients who began CSF therapy beyond 1 week after chemotherapy completion (average, 16.8 days) likely represents those patients who were begun on CSF support at nadir, with some patients started in the inpatient setting. This practice is, in most circumstances, at variance with current guidelines. One of the most striking findings is that primary prophylaxis is quite common, with one half of patients receiving their initial CSF in cycle 1. This represents a prominent deviation from the ASCO guidelines, in which initiation of CSFs at the start of chemotherapy is recommended only in regimens for which the incidence of febrile neutropenia is greater than 40% (this represents only a small fraction of all chemotherapy regimens10 ) or in the case of special circumstances.2 Diagnosis and regimen selection as well as patient-specific factors likely influence these decisions regarding primary prophylaxis. On the other hand, we found in our study that few oncologists are initiating secondary prophylaxis after an episode of afebrile neutropenia. In our study, after an episode of afebrile neutropenia, only 7% of patients who were not receiving growth factors had CSF initiated during the next cycle. This represents an important area in which clinical practice is quite consistent with present guideline recommendations. The comparison of G-CSF and GM-CSF must be interpreted with caution. As a retrospective analysis, we cannot assume that the patients or the chemotherapy regimens used in these groups were in any way comparable. Retrospectively, we found no substantive difference between the two groups of patients when compared according to distribution of diagnosis, chemotherapy regimen used, age, or phase of treatment (data not shown). We have included these comparative data because of the large number of patients involved, their internal consistency, and the absence of any previous or current randomized controlled trials comparing these two agents in the clinical settings included in this article. One recent randomized study11 evaluated G-CSF versus GM-CSF as treatment for chemotherapy-induced afebrile neutropenia. However, no effort was made to ensure comparability of patients in the two arms, and current guidelines do not recommend CSF use in this setting.2 Clearly, a randomized controlled clinical study is the only way to definitively determine whether these two agents are comparable in terms of efficacy or toxicity. The data on practice variation are consistent with other studies that show wide ranges in application of a variety of cancer treatments.12,13 This also demonstrates that propagation of expert guidelines in national journals or within organizations is only a first and clearly insufficient step in effecting change in practice patterns. Our comparison with the ASCO surveys4 demonstrates the inadequacy of that format as a means of assessing actual physician behavior. We feel that our data more accurately represent what happens in the community setting, outside of the restrictions of clinical trial participation. There are several shortcomings to this study. Because the computer system is restricted to the oncology office, CSF usage in the inpatient setting is unknown. Therefore, we cannot comment on hospital-based use, such as treatment of febrile neutropenia. Also, these data were collected in a data warehouse and were retrospectively analyzed, with the possible attendant biases that accompany this type of review. Although the data were entered at the time of patient encounter, the types of analyses were selected after all of the data had been collected. Our efforts, given this information, are now focused on two interventions designed to assist physicians in clinical decision making. First, individual feedback is provided to all oncologists comparing practitioner, practice, and organization-wide clinical activities. Previous studies have shown that this type of personalized information can impact physician behavior.14 Second, we are in the process of replacing the OPUS Matrix System, which focuses on chemotherapy ordering and administration, with a more sophisticated full-spectrum electronic medical record developed by iKnowMed Systems (Berkeley, CA). It presents computer-generated, guideline-based, patient-specific treatment recommendations to the physician at the time of treatment ordering and solicits reasons for guideline deviations. We can now use the data from this study to integrate specific physician reminders and alerts into the electronic medical record to encourage greater adherence to the ASCO CSF guidelines. Again, previous studies have shown that clinical decisions can clearly be influenced by presenting the physician with relevant information at the time of clinical encounter.15 However, development and implementation of sophisticated information systems is complex and costly and requires a strong institutional commitment.16 In summary, this study demonstrates the central role that information technology can play in the assessment of treatment of patients with cancer. Only through the integration of systems such as this can we gather the data that are necessary to reliably know how our patients are being treated and how we can continue to improve their care.
We thank all of the OnCare nurses and physicians who use the Matrix System; without their work, this publication would not have been possible.
1. American Society of Clinical Oncology: Recommendations for the use of hematopoietic colony-stimulating factors: Evidenced-based clinical practice guidelines. J Clin Oncol 12:2471-2508, 1994
2.
Ozer H, Miller LL, Schiff CA, et al: Update of recommendations for the use of hematopoietic colony stimulating factors: Evidence based clinical practice guidelines. J Clin Oncol 14:1957-1960, 1996 3. Bennett CL, Smith TJ, Weeks JC, et al: Use of hematopoietic colony stimulating factors: The American Society of Clinical Oncology survey. J Clin Oncol 14:2511-2520, 1996[Abstract]
4.
Bennett CL, Weeks JA, Somerfield MR, et al: Use of hematopoietic colony-stimulating factors: Comparison of the 1994 and 1997 American Society of Clinical Oncology Surveys regarding ASCO clinical practice guidelines. J Clin Oncol 17:3676-3681, 1999 5. Lieschke GJ, Fotte MA, Morstyn G: Hematopoietic growth factors in cancer chemotherapy, in Longo DL, Chabner BA (eds): Cancer Chemotherapy and Biological Response Modifiers. New York, NY,Elsevier Science, 1997, pp 363-389 6. Glaspy JA: Economic outcomes associated with the use of hematopoietic growth factors. Oncology 9:93-105, 1995 (suppl)[Medline]
7.
Lyman GH, Lyman CG, Sanderson RA, et al: Decision analysis of hematopoietic growth factor use in patients receiving cancer chemotherapy. J Natl Cancer Inst 85:488-493, 1993
8.
Armitage JO: Emerging applications of recombinant human granulocyte-macrophage colony-stimulating factor. Blood 92:4491-4508, 1998 9. Lyman GH, Balducci L: A cost analysis of hematopoietic colony stimulating factors. Oncology 9:85-91, 1995 (suppl)[Medline] 10. Smith TJ: Role of granulocyte- and granulocyte-macrophage colony-stimulating factors in clinical practice: Balancing clinical and economic concerns. ASCO Educational Book, spring:275–280, 1999 11. Beveridge RA, Miller JA, Kales AN, et al: A comparison of efficacy of sargramostim (yeast-derived RhuGM-CSF) and filgrastim (bacteria-derived RhuG-CSF) in the therapeutic setting of chemotherapy-induced myelosuppression. Cancer Invest 16:366-373, 1998[Medline] 12. Nattinger AB, Gottlieb MS, Veum J, et al: Geographic variation in the use of breast-conserving treatment for breast cancer. N Engl J Med 326:1102-1107, 1992[Abstract] 13. Lu-Yao GL, McLerran D, Wasson J, et al: An assessment of radical prostatectomy time trends, geographic variation and outcomes. JAMA 296:2633-2636, 1993 14. Davis DA, Thomson MA, Oxman AD, et al: Changing physician performance: A systematic review of the effect of continuing medical education strategies. JAMA 274:700-705, 1995[Abstract]
15.
Hunt DL, Haynes RB, Hanna SE, et al: Effects of computer-based clinical decision support systems on physician performance in patient outcomes: A systematic review. JAMA 280:1339-1346, 1998 16. Mighion K, Gesme DH, Rifkin RM, et al: Growth of oncology physician practice management companies. Cancer Invest 17:362-370, 1999[Medline] Submitted July 6, 1999; accepted December 15, 1999. This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2000 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
2 test. Homogeneity of odds ratios was tested with a Breslow-Day test. The pooled odds ratio expresses the probability of individual events occurring with the use of GM-CSF versus G-CSF. 
