Antifungal prophylaxis in chemotherapy-associated neutropenia: a retrospective, observational study

Background In August 2002, the antifungal prophylaxis algorithm for neutropenic hematology/oncology (NHO) patients at the Medical Center was changed from conventional amphotericin (AMB) to an azole (AZ) based regimen (fluconazole [FLU] in low-risk and voriconazole [VOR] in high-risk patients). The aim of our study was to compare outcomes associated with the two regimens, including breakthrough fungal infection, adverse drug events, and costs. Methods Adult, non-febrile, NHO patients who received prophylactic AMB from 8/01/01-7/30/02 or AZ from 8/01/02-7/30/03 were retrospectively evaluated. Results A total of 370 patients (AMB: n = 181; AZ: n = 216) associated with 580 hospitalizations (AMB: n = 259; AZ: n = 321) were included. The incidence of probable/definite breakthrough Aspergillus infections was similar among regimens (AMB: 1.9% vs AZ: 0.6%; p=0.19). A greater incidence of mild/moderate (24.7% vs. 5.3%; p < 0.0001) and severe renal dysfunction (13.5% vs. 4.4%; p < 0.0012) was observed with AMB. In contrast, patients treated with VOR were found to have an increased rate of severe hepatic toxicity (32.5%) compared with patients treated with either AMB (22.6%) or FLU (21.4%) (p = 0.05). While the AZ period was associated with a >$9,000 increase in mean total costs/hospitalization, the mean acquisition cost associated with AZ was only $947/hospitalization more than AMB. Conclusion While an AZ-based regimen is associated with increased cost, the reduced rate of nephrotoxicity and availability of oral dosage forms, suggests that azoles be used preferentially over AMB. However, an increased rate of severe hepatic toxicity may be associated with VOR.


Background
Invasive fungal infection due to Candida and Aspergillus is associated with increased morbidity and mortality in neutropenic hematology and oncology (NHO) patients.
Amphotericin B (AMB) is a broad-spectrum antifungal agent used for prophylaxis of fungal infections in neutropenic hematology/oncology patients. However, conventional AMB is associated with an increased risk of infusion-related reactions and renal toxicity. Prospective, randomized, controlled trials have demonstrated that fluconazole (FLU) and voriconazole (VOR) are safe and effective alternatives to AMB in febrile, neutropenic patients and fluconazole has been found to be an effective prophylactic agent in neutropenia [1][2][3][4]. The Centers for Disease Control, Infectious Disease Society of America, and American Society of Blood and Bone Marrow Transplantation currently recommend FLU for the prevention of candidiasis for patients undergoing hematopoietic stem cell transplantation [5]. Of concern, FLU has inadequate activity versus Aspergillus spp. Although VOR has adequate activity versus Aspergillus spp, it is currently not approved by the Food and Drug Administration for use in febrile neutropenia. However, Walsh et al. (2002) demonstrated that VOR was superior to liposomal AMB in reducing documented breakthrough fungal infections (p = 0.02), severe infusion-related toxicity (p < 0.01), and renal toxicity (p < 0.001) in patients with febrile neutropenia [2]. Based on the results of this study, in August 2002, the University of California at San Francisco (UCSF) antifungal prophylaxis regimen was changed from conventional AMB to an azole-based regimen that included FLU and VOR. The objectives of this study were to compare the rate of breakthrough fungal infection, adverse drug events and costs associated with these prophylactic regimens.

Eligibility criteria
The UCSF Committee on Human Research approved the study. All patients were treated at the UCSF Medical Center. Adult neutropenic (defined as an absolute neutrophil count < 500 cells/mm 3 ) hematology/oncology patients who received conventional AMB prophylaxis from August 1, 2001 through July 30, 2002, or who received FLU or VOR prophylaxis from August 1, 2002 to July 30, 2003, were included. During the AZ period, lowrisk patients, defined as leukemics or those receiving autologous transplant, received prophylaxis with FLU. High-risk patients, defined as those receiving allogeneic stem cell transplants, received prophylaxis with VOR. Patients were excluded if they received azoles or AMB for reasons other than prophylaxis.

Administration of Study Medications
In accordance with the UCSF antifungal prophylaxis algorithm, from August

Study Design
This investigation was a retrospective, observational, cohort study. Non-blinded reviews of patients' electronic files at UCSF, the pharmacy acquisition cost database, and the medical center cost accounting system were conducted. Collected information included: gender, age, diagnosis, cost of hospitalization (including pharmacy, chemistry laboratory, and hemodialysis), length of hospital stay, renal toxicity, hepatotoxicity, breakthrough fungal infections, hemodialysis (HD) requirements, reason for HD, and modification of antifungal prophylaxis. Duration of antifungal prophylaxis was calculated by the number of days from the start of therapy to discharge.

Efficacy assessment
The definitions of possible, probable, and definite breakthrough fungal infections were based upon clinical consensus guidelines on opportunistic fungal infections in cancer patients and stem cell recipients [6].

Safety assessment
Renal and hepatic toxicity were defined as per previous published trials of AMB, FLU, and VOR [2][3][4]7,8]. Mild to moderate renal toxicity was defined as >0.4 mg/dL increase in serum creatinine (Scr). Severe renal dysfunction was defined as at least a doubling of the Scr level (or an increase of at least 1 mg/dL if the base line was above upper normal limits).
Hospitalizations associated with acute HD or continuous veno-venohemofiltration (CVVH) were identified using billing codes. The patient's electronic files were further reviewed to determine etiology of renal failure and confirm the mode of dialysis.
Mild to moderate hepatic toxicity was defined as a tripling of the baseline value of serum aspartate aminotransferase (AST), or serum alanine aminotransferase (ALT), or serum alkaline phosphatase (Alk Phos) or a 1.5-fold increase in serum total bilirubin during therapy. Severe hepatic toxicity was defined as at least a five-fold increase in the baseline value of AST, ALT, or Alk Phos or a three-fold increase in serum total bilirubin during therapy.

Cost Analysis
Total hospital costs represent the UCSF cost of providing service and do not reflect charges or reimbursement. TSI (Transition System Inc., Eclipsys) was used to determine total, pharmacy and laboratory costs for all patients.

Statistical Analysis
Due to multiple hospitalizations for patients, continuous variables (age, length of stay, duration of therapy, baseline and peak lab values [Scr, AST, ALT, Alk Phos, T bili], costs) were analyzed using mixed effects logistic regression. Some of these variables (length of stay, duration of therapy, costs) were then log transformed to improve the distribution. These models controlled for the number of hospitalizations and include fixed effects for each medication and for each period.
Similarly, discrete variables (rates of infection, renal toxicity, hepatic toxicity) were analyzed using random effects logistic regression. These models controlled for the number of hospitalizations and include fixed effects for each medication and for each period.
Multi-variable analyses were conducted for the outcome of total costs using mixed effects regression models that controlled for age, diagnoses, therapy (allogeneic transplant, autologous transplant, chemotherapy, other diagnoses), and duration of antifungal prophylaxis. The cost variables were then log transformed to improve the distribution. Multi-variable analyses were also conducted for the outcome of all fungal infection using random effects regression models that controlled for age, diagnoses, therapy (allogeneic transplant, autologous transplant, chemotherapy, other diagnoses), and duration of antifungal prophylaxis. These models include fixed effects for each medication and for each period.
Statistical significance was defined as a P value of < 0.05.

Patient Characteristics
A total of 370 patients associated with 580 hospitalizations were included in the analysis (Table 1). During the AZ period, FLU (low-risk patients) was used in 237 (74%)

Patients Treated in Both the AMB and AZ Periods
Twenty-seven patients, representing 82 hospitalizations, were treated in both periods. These patients were determined to have similar rates of renal dysfunction, hepatic dysfunction, and breakthrough fungal infections as those patients solely treated in a single study period.

Cost Analysis
When the cost of the AZ period was compared to the AMB period, a $9,128/hospitalization increase in the mean total cost, $9,390/hospitalization increase in the mean pharmacy costs, and $68/hospitalization increase in mean laboratory costs was observed (Tables 3, 4). Multivariable analyses confirmed significantly less cost associated with AMB period (95% CI 0.65 -0.78; p < 0.0001).
Of note, caspofungin, an expensive agent by acquisition cost, was available solely during the AZ period. When the cost of caspofungin was removed from the analysis for the AZ period, the mean increase in total cost and pharmacy cost decreased to $8,916 and $9,090, respectively. While the increase in pharmacy costs was substantial between periods, the increase in mean acquisition cost of antifungals increased by only $947/hospitalization during the AZ period. In addition to the increased cost associated with FLU and VOR, a $157,490 increased total cost associated with greater use of caspofungin was observed during the AZ period. When this cost of caspofungin was removed from the analysis, the increased mean antifungal cost of the AZ period further decreased to $476/hospitalization. Of note, severe renal dysfunction was associated with a $20,465 (95% CI $10,458 to $33,729; p < 0.0004) increased total cost and $5,973 (95% CI $2199 to $11,400; p = 0.03) increased pharmacy cost per hospitalization. Compared to those patients without hepatic toxicity, patients with severe hepatic toxicity had an associated increased total hospital cost of $35,543 (95% CI $30,409, to $41,499; p < 0.0001) and increased pharmacy cost of $12,411 (95% CI $10,178 to $15,111; p < 0.0001).

Discussion
The current study was designed to compare patient outcomes and costs associated with AZ or AMB prophylaxis in NHO patients. While retrospective in design, the current investigation is the largest with respect to the evaluation of VOR for primary prophylaxis in NHO patients.
There are several limitations to the study. As with any retrospective, observational cohort study, the influence of confounding factors is a concern. An additional challenge was analyzing those patients who received multiple courses of antifungals. Using mixed or random effects logistic regression models that controlled for the number of hospitalizations and includes fixed effects for each medication and each period, this bias was statistically minimized. The subgroup analyses of those patients with multiple admissions, however, suggest that multiple hospitalizations did not correlate with worse outcomes.
Not surprisingly, the incidence of mild/moderate and severe renal dysfunction was significantly higher in the AMB period, consistent with other reports [1,8]. Notably, renal dysfunction was associated with significantly increased costs, independent of the study period.
While the nephrotoxic potential of conventional amphotericin is well recognized, the need for hemodialysis was found to be infrequent.
The rate of severe hepatic toxicity was not significantly different between the AMB and AZ periods. However, patients treated with VOR had a significantly greater rate of severe hepatic toxicity. Our findings regarding hepatic toxicity associated with VOR differ from those reported in previous prospective, controlled clinical studies; it is possible that the increased rate may be due to confounding factors, including concomitant use of hepatotoxic agents [1][2][3][4]7]. Nevertheless, VOR was associated with increased rate and severity of liver dysfunction.
While not statistically significant, the incidence of probable and definite breakthrough fungal infections was greater during the AMB period. However, it is interesting to note that all breakthrough Candida infections and 4 of the 5 cases of fatal possible breakthrough Aspergillus infections occurred in those patients who received FLU prophylaxis.
The AZ period was associated with an increase of >$9,000/ hospitalization in the mean total and pharmacy costs. These increases in costs may represent the increased rate of allogeneic transplants conducted in significantly older patients treated during the AZ period. More recent improvement in supportive care medications have allowed for increased allogeneic and autologous transplantation in an older population of patients compared to the past. Despite the substantial increase in hospitaliza-  tion costs, the increase in costs directly associated with antifungal acquisition represented a small percentage of this increase in cost. Costs included in this study do not reflect the charges or the actual reimbursement associated with any of the costs. It should be noted the calculated cost to reimbursement ratio was similar during the two time periods (AZ: 1.196 and AMB: 1.153).

Conclusion
While an AZ-based prophylactic regimen is associated with increased cost, the reduced rate of nephrotoxicity and availability of oral dosage forms, suggests they be used preferentially over AMB. While the triazoles may offer certain advantages, an increased rate of hepatic toxicity may be associated with VOR when compared with AMB or FLU.