|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2000 American Society for Clinical Oncology Time to Clinical Response: An Outcome of Antibiotic Therapy of Febrile Neutropenia With Implications for Quality and Cost of CareFrom the The University of Texas M.D. Anderson Cancer Center, Houston, TX. Address reprint requests to Linda Elting, DrPH, Department of Health Services Research, The University of Texas M.D. Anderson Cancer Center, 1515 Holcombe Blvd, Box 40, Houston, TX 77030; email lelting{at}mdanderson.org
PURPOSE: To determine whether antibiotic regimens with similar rates of response differ significantly in the speed of response and to estimate the impact of this difference on the cost of febrile neutropenia. METHODS: The time point of clinical response was defined by comparing the sensitivity, specificity, and predictive values of alternative objective and subjective definitions. Data from 488 episodes of febrile neutropenia, treated with either of two commonly used antibiotics (coded A or B) during six clinical trials, were pooled to compare the median time to clinical response, days of antibiotic therapy and hospitalization, and estimated costs. RESULTS: Response rates were similar; however, the median time to clinical response was significantly shorter with A-based regimens (5 days) compared with B-based regimens (7 days; P = .003). After 72 hours of therapy, 33% of patients who received A but only 18% of those who received B had responded (P = .01). These differences resulted in fewer days of antibiotic therapy and hospitalization with A-based regimens (7 and 9 days) compared with B-based regimens (9 and 12 days, respectively; P < .04) and in significantly lower estimated median costs ($8,491 v $11,133 per episode; P = .03). Early discharge at the time of clinical response should reduce the median cost from $10,752 to $8,162 (P < .001). CONCLUSION: Despite virtually identical rates of response, time to clinical response and estimated cost of care varied significantly among regimens. An early discharge strategy based on our definition of the time point of clinical response may further reduce the cost of treating non–low-risk patients with febrile neutropenia.
SINCE THE EARLY 1960s, significant strides have been made in the treatment of febrile neutropenia.1 New classes of antibiotics, including antipseudomonal penicillins, third-generation cephalosporins, carbapenems, and quinolones, have been developed. Some are effective as single agents. These advances, combined with insights into the use of empiric antibiotic therapy,2 combination regimens,3,4 and broad-spectrum single-agent regimens,5 now result in response rates to initial therapy that exceed 70% in many clinical trials. Today, in contrast to the high mortality rates of the early 1960s, fewer than 10% of patients with febrile neutropenia die from their infections. These developments have resulted in a variety of therapeutic options of similar safety and efficacy—so similar, in fact, that many clinical trials of empiric therapy report little difference among regimens. With only a few notable exceptions,5 within-class comparisons of antibiotics and comparisons of different multiple-drug regimens only occasionally produce improvements that reach clinical or statistical significance, despite adequate statistical power.6 However, our clinical experience suggests that significant benefits, in terms of the speed of response, are derived from some regimens that are not reflected in the commonly reported results of clinical trials. During these four decades, other major changes have occurred that influence the choice of antibiotic regimens. In the early 1960s, prevention of infection-related deaths among patients with cancer was the primary concern. As the antibiotic armamentarium expanded to reduce this concern, pressure to control the growth of medical costs increased. Outpatient treatment of patients at low risk for serious medical complications has been explored recently. Such treatment seems to afford an opportunity to reduce the cost of care by avoiding hospitalization in the majority of patients classified as low-risk.7-10 Although it may not be possible to avoid hospitalization entirely in non–low-risk patients, a step-down strategy that features early discharge at the time of clinical response could reduce the number of days of hospitalization required during episodes of febrile neutropenia.6 These two issues (perceived rapid response from some regimens and potential cost reduction through early discharge) set the stage for this study, in which we examined two hypotheses. First, we hypothesized that antibiotic regimens with virtually identical rates of response may differ significantly in the speed of response. Second, we hypothesized that differences in the speed of response could have a significant impact on the cost of care.
To test these hypotheses, we pooled data collected prospectively during clinical trials. We chose episodes of febrile neutropenia that were treated with either of two agents that have provided the backbone for single- and multidrug regimens for more than a decade. However, this is not a randomized trial that compares these two agents, despite the fact that the data are derived from such trials. Nor is it a formal meta-analysis, although data are pooled from multiple studies. Comparison of the two antibiotics is not among the goals of this report. In fact, they were chosen because they are used more or less interchangeably by many clinicians. Therefore, to avoid focusing on the specific agents, they were coded and reported as A and B throughout the study. Specifically, 488 episodes of febrile neutropenia were included from all six unblinded clinical trials of the two agents conducted at The University of Texas M.D. Anderson Cancer Center between 1981 and 1994.5,11-15 Because the speed of clinical response is affected by site of infection, an analysis stratified by site of infection was planned. To provide maximum power for the subset analysis by site of infection, all cases of bacteremia or pneumonia that were treated with drug A or drug B during the six studies, with or without vancomycin (V), were included. A random sample of episodes of fever of unknown origin was selected, such that the power to detect a difference in response rate of 20% between regimens was 80% at a level of significance of P = .05. Eleven patients (2%) who were receiving corticosteroids or other antipyretics were excluded from the analysis because of their effect on temperature, the principal outcome of interest for this study. Eight patients (2%) with polymicrobial bacteremias were excluded because they constituted a heterogeneous group from which it would have been difficult to generalize results. Clinical data, microbiology data, response to initial therapy, and ultimate outcome of therapy were extracted from the Infectious Diseases Clinical Trials Database. These data were collected prospectively during the original clinical trials according to standard data collection protocols using standardized definitions. Additional data were abstracted from patients’ medical records, including vital signs taken every 4 hours during therapy.
Outcomes In the alternative analysis, an intermediate outcome, time to clinical response, was defined as the time between the onset of antibiotic therapy and the point at which clinical response had occurred. Because no single definition for the time point of clinical response is widely accepted, we first evaluated four alternative time points (alone and in combination) for their value in predicting response to antibiotic therapy as defined above. The four time points were as follows: (1) the point at which the temperature first fell below 37.8°C, (2) the point at which three consecutive 4-hourly temperatures less than 37.8°C had occurred, (3) the point at which the temperature had remained below 37.8°C for 24 hours, and (4) the point at which the patient reported subjective response (felt better).
Outcome Modifiers
Resource Use and Cost Estimates of the cost of hospitalization were obtained by applying an average daily cost of hospitalization for febrile neutropenia ($1,002) to the total number of days of hospitalization for febrile neutropenia for each patient.17 These average daily costs were obtained from separate patients who received either A or B as standard therapy off-protocol, not as a part of clinical trials, and they include only those costs attributable to the treatment of infection. They include the costs of diagnostic imaging and microbiology as well as hotel (room) costs. The costs of diagnostic testing or treatment of the underlying cancer incurred during the hospitalization were not included in the total cost estimate. Estimates of the cost of hypothetical home administration of antibiotics were similarly obtained by applying an average daily cost ($133) to the total number of days of home care. We assumed that the intravenous (IV) antibiotic regimen begun in the hospital would be continued at home for the same duration. The average daily costs of hospital and home care were assumed to be the same across the four antibiotic regimens.17 Professional fees and the costs of blood products, growth factors, catheter removal, and adverse reactions were not available for this analysis. The costs of antibiotics were computed from the current average wholesale price for each agent (drug B = $14.22 per gram; drug A = $27.76 per 500 mg; V = $7.80 per 500 mg) and the daily dose and schedule. An additional cost of $25 for each infusion episode was added to cover the cost of IV tubing and fluids and preparation and administration of the dose.
Statistical Considerations This study involved primary analysis of pooled data from several prospective trials that were conducted in the past. The outcomes of interest, time to clinical response and cost, were neither analyzed nor collected for the initial analyses. Therefore, primary analyses of the pooled data rather than meta-analytic techniques were used. The original trials were conducted using identical clinical protocols and identical, standardized definitions of outcomes and covariates. However, pooling of these data was attempted only after comparisons of the characteristics of subjects and infections across all studies revealed no unexpected, substantive differences (Table 1). Differences in causative organisms and in the depth and duration of granulocytopenia were anticipated because of well-documented changes in spectra of infections and patterns of care over time. Therefore, the prevalence of these factors was assessed in the four treatment groups.
Four groups were compared: A, B, A + V, and B + V. Statistical tests compared similar regimens. A alone was compared with B alone; A + V was compared with B + V. Differences between categorical variables were tested with univariate two-tailed  2 tests. Differences between continuous variables were compared using the Mann-Whitney test. The time to response was also evaluated by Kaplan-Meier analysis, and differences were tested using the Breslow-Day statistic, which emphasizes observations early in the follow-up period, the time during initial antibiotic therapy. Statistical tests were computed using BMDP Dynamic software (BMDP Statistical Software, Inc, Los Angeles, CA).
After we excluded 19 episodes for the previously mentioned reasons, 488 episodes of febrile neutropenia in 466 patients remained for analysis. These subjects would generally be classified as non–low risk for various reasons. Their durations of granulocytopenia exceeded 1 week in a majority of cases. A significant proportion (48%) of these patients were hospitalized for other reasons at the time their infections developed, hematologic malignancies were common (44%), and 15% had undergone bone marrow transplantation. Fifty-six percent of patients with acute leukemia underwent reinduction chemotherapy after relapse, and 46% with solid tumors had widely disseminated metastases.
Comparability of Treatment Groups
Time Point of Clinical Response: Development of a Working Definition At the time of analysis and reporting of the results of a clinical trial, the time point of clinical response can be determined for each patient in retrospect, based on the entire fever curve. However, that is not possible in clinical decision making. Therefore, we tested four potential time points for their accuracy in classifying patients prospectively, as would be done in clinical practice. For this test, the gold standard was the response to the antibiotic regimen being received at the time the evaluation was made. A single normal temperature was highly sensitive (99%), but its specificity was unacceptably low (8%) (Table 3). Twenty-eight percent of episodes classified as clinical responses subsequently failed the antibiotic regimen being administered at that time. Furthermore, 29 of 32 episodes that were ultimately fatal would have been classified as clinical responses. If an early-discharge strategy had been based on this definition, 29 patients whose deaths occurred during antibiotic therapy and as a result of infection would have been sent home inappropriately.
The addition of patient-reported subjective response significantly improved the specificity of a single normal temperature, from 8% to 68%, with little loss of sensitivity. However, 13% of episodes classified as responses failed that antibiotic regimen, and nine of 32 ultimately fatal episodes would have been classified as clinical responses (Table 3). Although the 24 hours afebrile plus subjective response definition does not provide a significant improvement in overall specificity over the three consecutive normal temperatures definition, they differ in one important aspect. With the former definition, all 32 fatal episodes would have been classified properly as clinical failures compared with only 23 (72%) of 32 with the latter definition. We consider this difference of sufficient clinical importance to justify choosing the former definition for the remainder of the analysis in this study.
Outcomes of Antibiotic Therapy: Comparison of Conventional and Alternative Outcomes
Response occurred a median of 2 days sooner during episodes treated with A compared with episodes treated with B (P = .003), whereas those treated with A + V responded 1 day sooner than those treated with B +V (P = .09). After 72 hours of therapy, 33% of episodes treated with A and only 18% of those treated with B had responded to therapy (P = .01). Kaplan-Meier analysis of the time to clinical response showed similar results (Fig 1). The time to clinical response curves for A-based regimens are significantly different (P = .007) but eventually converge, which reflects differing speeds of response but similar response rates.
The difference in time to response resulted in significantly shorter durations of antibiotic therapy and hospitalization for episodes treated with A compared with those treated with B (P < .04, both comparisons) (Table 6). The shorter durations of antibiotic therapy and hospitalization led to a significantly lower estimated total cost of actual care for A-based regimens compared with those based on B. Although the acquisition price of A exceeds that of B (a metric customarily used in formulary decisions), that difference is trivial compared with the cost of an additional hospital day.
With the hypothetical application of a strategy consisting of early discharge to home antibiotic therapy at the time of clinical response (subjective response plus 24 hours afebrile), the estimated median total cost of care ($8,162) would have been significantly decreased compared with the actual hospital-based care of these patients ($10,752; P < .001). Again, A-based regimens would have been significantly less costly than B-based regimens.
Recent randomized controlled clinical trials of empiric antimicrobial therapy of febrile neutropenia are notable in that significantly different response rates between treatment regimens are rarely demonstrated. Our study, which combined the results of several clinical trials, is no different. Response rates to the four regimens ranged from 63% to 77%, and none of the differences were statistically significant, despite large sample sizes after pooling. Based on this observation and on previous research, one might conclude that these regimens are interchangeable. However, striking differences among regimens emerge when the time to clinical response is examined. In this respect, A-based regimens were superior to B-based regimens. Although these results provide support for the notion that otherwise similar antibiotic regimens may differ in clinically significant outcomes, this was not a formal trial of specific antibiotic regimens. However, a recent randomized trial contains similar findings.18 Freifeld et al18 report that at 72 hours of therapy, 74% of patients receiving imipenem had defervesced compared with only 60% of those receiving ceftazidime (P = .02). In fact, 6% of patients receiving ceftazidime and 13% receiving imipenem (P = .03) were removed early from this randomized trial to participate in a separate study of step-down care with oral antibiotics. The mean time to defervescence for patients receiving imipenem alone was 3 days, compared with 4 days in those receiving ceftazidime, although this difference apparently did not reach statistical significance. However, the response rates to initial therapy were 79% and 80%, respectively, and by the end of the study, the overall response rates were also virtually identical (99% v 98%, respectively). Considered together, the results of these studies offer compelling evidence that the time to clinical response may differ significantly between regimens with similar response rates. The significantly shorter times to clinical response noted with A-based regimens underscore the importance of examining all clinically important outcomes in clinical trials. In this respect, design of antibiotic trials that involve patients with cancer may have lagged behind that of antineoplastic trials. Early trials of antineoplastic agents also focused on rate of response. However, as effective agents became more numerous (as is clearly the case with antibiotics), the rate of response was joined by outcome measures such as disease-free interval and duration of survival, and more recently, functional status, quality of life, and cost. Our study clearly demonstrates the value of such analyses in distinguishing between similarly effective antibiotic regimens in clinical trials; this measure should be included in guidelines for the design and analysis of such trials.19 Our estimates of the cost of the actual care delivered illustrate the association between differences in time to clinical response and corresponding differences in the cost of actual care. They also suggest opportunities for cost savings through step-down care. The concept of step-down care is well accepted in many areas of medicine. In the treatment of febrile neutropenia in both adults and children, early discontinuation of antibiotics (before neutrophil recovery) among patients with fever of unknown origin has been attempted with varying success.9,20-23 Step-down care has also been addressed in two recent trials that described the successful use of oral antibiotics among hospitalized patients with febrile neutropenia.24,25 We examined the potential of a stepped-down care strategy consisting of early discharge with continued antibiotics rather than early discontinuation of antibiotics. A hypothetical strategy consisting of inpatient IV antibiotic therapy until clinical response followed by stepped-down care at home was examined for its potential impact on both the quality and cost of care. Although the safety and efficacy of such an approach remain to be demonstrated in prospective clinical trials, some preliminary conclusions may be drawn from this study. Our results illustrate the potential cost savings that may be realized if safety issues can be resolved and a step-down, early-discharge strategy can be implemented. These savings may exceed 24% of the estimated median total cost per episode ($2,500). In addition to its impact on the financial bottom line, reducing the duration of hospitalization (which accounted for most of the cost savings) also has an impact on access to care. A step-down early-discharge strategy could free hospital beds for patients more likely to benefit from hospital-based care. The costs and savings reported in this article are estimates and, therefore, may not reflect the true cost of treating febrile neutropenia. Furthermore, the estimates were derived from clinical trials that require rigid adherence to structured protocols. The cost of care in general practice could be different. The cost of care could also be different because the design of this study precluded the collection of the costs of adverse events, blood products, and growth factors. These limitations could alter the differences among regimens substantially. For example, the toxicity profiles of the two antibiotics studied are different. More frequent adverse events with A-based regimens could alter the differences in cost that were observed. However, the differences between usual care and step-down care for patients receiving the same regimen and between times to clinical response should be unaffected. Also of note, the hypothetical cost savings are dependent on the discharge of patients who have achieved a clinical response to antibiotic therapy. Many patients would be ineligible for early discharge for reasons other than their response to antibiotic therapy. Therefore, the greatest impact of a step-down strategy would be realized among outpatients who were admitted solely for the treatment of febrile neutropenia. Furthermore, the financial benefits also depend on the availability of resources to support home antibiotic therapy. If these are unavailable or if reimburse-ment issues preclude their widespread use, such a strategy, although clinically feasible, would be impractical. A question of critical importance to the quality of care is, When can a non–low-risk patient with febrile neutropenia be safely discharged on antibiotics? In testing a number of potential definitions of this time point, the safety issue is abundantly clear. Definitions that seemed to provide adequate sensitivity and specificity overall nevertheless misclassified patients who ultimately died from their infections. Although our definition avoided this pitfall in this data set, the impact of this strategy on safety cannot be assumed. The definition has not been validated in another data set, and as pointed out in a recent editorial,26 such innovations in therapy must be tested prospectively before they can safely be introduced into practice. Finally, consideration of cost and resource use aside, antibiotic regimens that produce more rapid clinical response must be superior from a patient’s perspective. Quality of life was not studied formally, but the superiority of more rapid clinical response is self-evident when the subjective component of our definition of time to clinical response is considered. All other things being equal, feeling better sooner certainly must be better than feeling better later. We conclude that the time to clinical response is an important consideration in the choice of antibiotic regimens to treat febrile neutropenia in patients with cancer. We suggest that this critical outcome be measured in future clinical trials that compare regimens and that the definition for the outcome measure developed during this study be used because of its high specificity and its clinical significance. On the basis of these data, we propose that a strategy for step-down care of non–low-risk patients with febrile neutropenia be tested in prospective clinical trials. Successful implementation of these strategies may have far-reaching implications for the cost of care.
The two antibiotics examined in this report were imipenem (A) and ceftazidime (B). As previously indicated, this was not a formal trial comparing the efficacy of these two drugs, although every attempt was made to ensure the comparability of the samples. For that reason, great caution should be used in drawing conclusions from these results about the comparative efficacy of these two agents.
Supported in part by a grant from Merck & Co, Inc, Whitehouse Station, NJ.
Presented in part at the Second International Febrile Neutropenia Meeting, Brussels, Belgium, December 14-16, 1995, and the Annual Meeting of the Multinational Association for Supportive Care in Cancer, St Galen, Switzerland, February 26 through March 1, 1997.
1. Bodey GP: Antibiotics in patients with neutropenia. Arch Intern Med 144: 1845-1851, 1984[Abstract] 2. Schimpff S, Satterlee W, Young V, et al: Empiric therapy with carbenicillin and gentamicin for febrile patients with cancer and granulocytopenia. N Engl J Med 284: 1061-1065, 1971 3. Schimpff S: Overview of empiric antibiotic therapy for the febrile neutropenic patient. Rev Infect Dis 7: S734-S740, 1985 4. Pizzo P: Granulocytopenia and cancer therapy: Past problems, current solutions, future challenges. Cancer 54: 2649-2661, 1984[Medline] 5. Rolston K, Berkey P, Bodey G, et al: A comparison of imipenem to ceftazidime with or without amikacin as empiric therapy in febrile neutropenic patients. Arch Intern Med 152: 283-291, 1992[Abstract] 6. Hughes W, Armstrong D, Bodey G, et al: 1997 Guidelines for the use of antimicrobial agents in neutropenic patients with unexplained fever. Clin Infect Dis 25: 551-573, 1997[Medline] 7. Talcott J, Finberg R, Mayer R, et al: The medical course of cancer patients with fever and neutropenia: Clinical identification of a low-risk subgroup at presentation. Arch Intern Med 148: 2561-2568, 1988[Abstract] 8. Talcott J, Siegel R, Finberg R, et al: Risk assessment in cancer patients with fever and neutropenia: A prospective, two-center validation of a prediction rule. J Clin Oncol 10: 316-322, 1992[Abstract] 9. Rubenstein E, Rolston K, Benjamin R, et al: Outpatient treatment of febrile episodes in neutropenic cancer patients: A randomized prospective comparison of oral ciprofloxacin and clindamycin to intravenous aztreonam and clindamycin. Cancer 71: 3640-3646, 1993[Medline] 10. Malik I, Khan W, Aziz Z, et al: Self-administered antibiotic therapy for chemotherapy-induced, low-risk febrile neutropenia in patients with nonhematologic neoplasms. Clin Infect Dis 19: 522-527, 1994[Medline] 11. Fainstein V, Bodey GP, Elting L, et al: A randomized study of ceftazidime compared to ceftazidime and tobramycin for the treatment of infections in cancer patients. J Antimicrob Chemother 12: S101-S110, 1983 12. Bodey G, Elting L, Jones P, et al: Imipenem/cilastatin therapy of infections in cancer patients. Cancer 60: 255-262, 1987[Medline] 13. Anaissie E, Fainstein V, Bodey G, et al: Beta-lactam regimens in febrile neutropenic cancer patients. Am J Med 84: 581-589, 1988[Medline] 14. Bodey G, Fainstein V, Elting L, et al: Beta-lactam regimens for the febrile neutropenic patient. Cancer 65: 9-16, 1990[Medline] 15. Raad I, Whimbey E, Rolston K, et al: A comparison of aztreonam plus vancomycin and imipenem plus vancomycin as initial therapy for febrile neutropenic cancer patients. Cancer 77: 1386-1394, 1996[Medline] 16. Elting L, Rubenstein E, Rolston K, et al: Outcomes of bacteremia in neutropenic cancer patients: Observations from two decades of epidemiologic and clinical trials. Clin Infect Dis 25: 247-259, 1997[Medline] 17. Cantor S, Rubenstein E, Elting L, et al: Economic evaluation of management strategies for low-risk febrile neutropenic cancer patients. Proc Am Soc Clin Oncol 16: 419a, 1997 (abstr 1498)
18.
Freifeld A, Walsh T, Marshall D, et al: Monotherapy for fever and neutropenia in cancer patients: A randomized comparison of ceftazidime versus imipenem. J Clin Oncol 13: 165-176, 1995 19. From the Immunocompromised Host Society: The design, analysis, and reporting of clinical trials on the empirical antibiotic management of the neutropenic patient. J Infect Dis 161: 397-401, 1990[Medline] 20. Rodriguez V, Burgess M, Bodey G: Management of fever of unknown origin in patients with neoplasms and neutropenia. Cancer 32: 1007-1012, 1973[Medline] 21. Pizzo P, Robichaud K, Gill F, et al: Duration of empiric antibiotic therapy in granulocytopenic patients with cancer. Am J Med 67: 194-200, 1979[Medline] 22. Mullen C, Buchanan G: Early hospital discharge of children with cancer treated for fever and neutropenia: Identification and management of the low-risk patient. J Clin Oncol 8: 1998-2004, 1990[Abstract]
23.
Tomiak A, Yau J, Huan S, et al: Duration of intravenous antibiotics for patients with neutropenic fever. Ann Oncol 5: 441-445, 1994
24.
Freifeld A, Marchigiani D, Walsh T, et al: A double-blind comparison of empirical oral and intravenous antibiotic therapy for low-risk febrile patients with neutropenia during cancer chemotherapy. N Engl J Med 341: 305-311, 1999
25.
Kern W, Cometta A, DeBock R, et al: Oral versus intravenous empirical antimicrobial therapy for fever in patients with granulocytopenia who are receiving cancer chemotherapy. N Engl J Med 341: 312-318, 1999
26.
Finberg R, Talcott J: Fever and neutropenia: How to use a new treatment strategy. N Engl J Med 341: 362-363, 1999 Submitted February 17, 2000; accepted June 16, 2000. 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
