Newer quinolone agents with broader in vitro gram-positive activity have recently emerged as possible candidates for antibacterial chemoprophylaxis regimens. Ofloxacin is a synthetic carboxyquinolone agent with in vitro activity against aerobic gram-negative bacilli similar to that of norfloxacin but with greater activity against gram-positive organisms, such as staphylococci and streptococci [10]. Early studies [11-15] suggested a promising role for this agent for antibacterial prophylaxis in neutropenic patients; however, gram-positive infections have remained a problem. Strategies to reduce gram-positive infection by combining prophylactic trimethoprim-sulfamethoxazole or quinolones with other gram-positive agents, such as erythromycin [16-18], roxithromycin [19], or the penicillins [15, 20, 21], have had variable success. Studies of prophylaxis with ofloxacin in neutropenic rats at the University of Manitoba [22] showed that adding rifampin to ofloxacin reduced both viridans streptococcal and staphylococcal bacteremia. Accordingly, the National Cancer Institute of Canada Clinical Trials Group designed an open, randomized, controlled phase III clinical trial to test the hypothesis that an increase in the gram-positive activity of an oral quinolone-based antibacterial prophylaxis regimen would have a clinically significant effect on the incidence of gram-positive infection in neutropenic patients with cancer receiving cytotoxic therapy.

Methods

Top Methods Results Discussion References

Patients

Patients between 18 and 70 years of age were eligible for inclusion in this trial if they 1) had acute myeloid leukemia or acute lymphocytic leukemia or had received an autologous bone marrow transplant; 2) were receiving cytotoxic therapy for primary remission-induction, reinduction for first relapse, postremission consolidation, or conditioning for bone marrow autografting; 3) had an absolute neutrophil count (segmented plus band neutrophils) of 0.5 x 10⁹/L or greater at study entry, with an expectation that severe neutropenia [neutrophil count less than 0.5 x 10⁹/L] would occur at 14 days or thereafter; 4) had a negative pregnancy test result [a criterion applicable to women of child-bearing age]; and 5) provided informed written consent. Patients were excluded if they were febrile (oral temperature more than 38 °C); had gastrointestinal conditions that might interfere with drug absorption; were hypersensitive to any of the study agents; had major renal (serum creatinine level more than 180 µmol/L), hepatic (liver function test results more than twice normal), or cardiovascular (uncontrolled hypertension or congestive failure) disorders; had received systemic antimicrobial therapy for more than 48 hours within 7 days before study entry; or had previously been entered in this trial.

Design

The study was an open, randomized, controlled phase III clinical trial. Before randomization, patients were stratified according to diagnosis and intent of cytotoxic therapy (primary remission-induction, reinduction, consolidation, or bone marrow autograft) to ensure an even distribution of patients among the study groups. The randomization list was generated by a computerized random-number generator. After patient eligibility was confirmed by telephone, assignments were allocated sequentially within the diagnostic grouping from the central office of the National Cancer Institute of Canada Clinical Trials Group. Equal numbers of patients were randomly allocated to receive norfloxacin, 400 mg orally every 12 hours; ofloxacin, 400 mg orally every 12 hours; or ofloxacin, 400 mg orally every 12 hours, plus rifampin, 300 mg orally every 12 hours. Antibacterial prophylaxis was begun on the first day of the cytotoxic regimen and was continued until antimicrobial agents were administered parenterally for suspected infection, the neutrophil count increased to 0.5 x 10⁹/L or greater, or the patient withdrew from the study for any reason. During the neutropenic period, infection was suspected if the patient's oral temperature increased to more than 38 °C on at least three occasions during a 12-hour period, if the patient's oral temperature increased to more than 39 °C on one occasion and was associated with chills or rigors, or if focal clinical signs of infection were present with a temperature greater than 38 °C on one occasion. When an infection was suspected, a history was obtained, a physical examination was done, blood samples were collected for culture, and, when appropriate, cultures were done on specimens obtained from sites of the suspected infection. Empiric antibacterial therapy with piperacillin and tobramycin was administered according to established guidelines [23].

Colonization Profiles

Serial surveillance cultures of the nasopharynx, oropharynx, and rectum were taken twice weekly during the study at one participating center (Winnipeg), as described elsewhere [24], to examine the effect of allocation on clearance and acquisition of aerobic gram-negative bacilli and on acquisition and colonization by opportunistic fungi. Acquisition was defined by the isolation of a microorganism not present in baseline surveillance cultures. Colonization was defined by isolation of the same species of microorganism from at least two sequential surveillance cultures at a given site, as described elsewhere [24].

Cytotoxic Therapy

Cytotoxic regimens were classified according to a cytotoxicity scale on the basis of cytarabine dosing: group 1, cytarabine-containing regimens with a total cytarabine dose greater than 2 g/m² body surface area (n = 24; 21.6%); group 2, cytarabine-containing regimens with a total cytarabine dose between 2 and 8 g/m² (n = 29; 26.1%); group 3, cytarabine-containing regimens with total doses greater than 8 g/m² [n = 23; 20.7%]; group 4, regimens for acute myeloid leukemia not based on cytarabine consisting of 1) etoposide at total doses of 0.5 g/m² or less plus idarubicin, mitoxantrone, or amacrine, 2) other high-dose etoposide-based regimens [more than 0.5 g/m²] plus carboplatin or melphalan, or 3) a high-dose cyclophosphamide regimen (50 to 60 mg/kg of body weight per day) (n = 15; 13.5%); and group 5, all other regimens (n = 20; 18%).

Outcome Measures

The primary end points were the incidence of suspected or documented infection and the spectrum of pathogens isolated. Infections were classified according to recommended guidelines [25] as microbiologically documented infections (bacteremic or nonbacteremic), clinical infections, or unexplained fevers. The secondary end points included adverse events related to study drugs, infections correlated by cytotoxic regimen, the time from allocation to suspected infection, the time to response to empiric antibacterial therapy, the time to infection according to cytotoxic regimen, and the colonization profiles.

Statistical Analysis

Categorical variables were analyzed using the Pearson chi-square statistic or the Fisher exact test when the expected cell sizes were small. For patients who had an infection, the time from the start of prophylaxis until the beginning of parenteral therapy was used as an end point representing time to infection. For patients without an infection, the time was censored at the last date of prophylaxis. Time-to-event analyses were done using the Kaplan-Meier product-limit method [26] and were compared using the log rank test. The prognostic relation of cytotoxic therapy to bacteremic infection was examined using a Cox regression procedure [27]. A P value less than 0.05 was considered statistically significant. All tests were two sided.

Results

Top Methods Results Discussion References

Eligibility and Evaluability

One hundred twenty-seven patients were enrolled from 14 centers between 1 July 1989 and 30 June 1992. The data set was frozen for analysis on 1 July 1994. Seven patients were ineligible because their neutrophil counts were less than 0.5 x 10⁹/L at randomization (n = 2), they previously received systemic antibiotic therapy (n = 3), their chemotherapy was insufficiently myelosuppressive (n = 1), or the underlying diagnosis was unacceptable according to the protocol (n = 1). Of the remaining 120 patients, 9 were inevaluable because they never received the study agents (n = 4) or because febrile events requiring parenteral antibiotics occurred within the first 24 hours after allocation (n = 5). Of the remaining 111 patients, 38 received norfloxacin, 38 received ofloxacin, and 35 received ofloxacin plus rifampin. All included patients had similar demographic characteristics (Table 1).

Adverse Events

Adverse events were graded as none, mild, moderate, and severe, and the relations of these events to the study regimens were designated as not related, possibly related, probably related, or definitely related. Gastrointestinal disturbance occurred in 28% of norfloxacin recipients, 51% of ofloxacin recipients, and 51% of ofloxacin plus rifampin recipients (p equals 0.07). The incidence of gastrointestinal disturbances that were considered probably or definitely related to the study agents was very low (2.8%) and was independent of treatment allocation. No other clinically important side effects were seen. Compliance, measured by dose intensity, was similar in all three groups (Table 1). Five patients died during this trial. Three (2.7%) died of their presenting infections, one from each treatment group. Two others died of causes unrelated to infection: intracranial hemorrhage in a norfloxacin recipient and ventricular tachyarrhythmia in an ofloxacin plus rifampin recipient.

Colonization Profiles

Serial surveillance culture studies were done in 31 patients. Forty-two aerobic gram-negative bacilli were isolated from baseline rectal surveillance cultures. All strains were cleared after a median of 5.5 days (range, 3 to 14 days). The clearance times were shorter in norfloxacin recipients and ofloxacin plus rifampin recipients than in ofloxacin recipients (median, 4 days for norfloxacin recipients, 5 days for ofloxacin plus rifampin recipients, and 7 days for ofloxacin recipients; P = 0.04 by log rank test). Six gram-negative isolates were transiently acquired only in single rectal cultures. Colonization with opportunistic yeasts occurred in at least one surveillance culture site in 22 patients (71%) and tended to be seen more often among ofloxacin plus rifampin recipients (10 of 11; 91%) than among norfloxacin recipients (5 of 7; 42%) or ofloxacin recipients (7 of 13; 23%) (P = 0.14). Colonization occurred more often in the oropharynx (58%) and rectum (61%) than in the nasopharynx (3%) (P < 0.001). Ofloxacin plus rifampin recipients acquired opportunistic yeasts earlier (median, 7 days) than did ofloxacin recipients (median, 13 days) or norfloxacin recipients (median, 22 days) (P < 0.01). We also found that ofloxacin plus rifampin recipients had earlier yeast colonization (median, 9 days) than did ofloxacin recipients (median, 13 days) or norfloxacin recipients (median, 28 days) (P = 0.11).

Infection

Table 2 shows the infections documented in our trial. Overall, 88 of 111 evaluable patients (79.3%) developed febrile episodes that required parenteral antibiotic therapy. Only three (2.7%) cases of gram-negative infection were seen: a soft-tissue infection caused by Bacteroides fragilis, bacteremia caused by Fusobacterium nucleatum, and a urinary tract infection caused by Escherichia coli that was susceptible to norfloxacin Table 3. Gram-positive infections developed in 27 patients (24%), accounting for 27 of 30 (90%) microbiologically documented infectious episodes.

The rates of microbiologically documented infections (occurring in 47% of norfloxacin recipients, 24% of ofloxacin recipients, and 9% of ofloxacin plus rifampin recipients; P < 0.001), bacteremic infections (occurring in 32%, 16%, and 6% of patients, respectively; P = 0.02), staphylococcal bacteremias (occurring in 24%, 13%, and 3% of patients, respectively; P = 0.03), and streptococcal bacteremias (occurring in 21%, 3%, and 3% of patients, respectively; P < 0.01) were reduced overall when norfloxacin was compared with ofloxacin and ofloxacin plus rifampin Table 2. This effect is shown in Figure 1 as the mean number of infections per patient. Ofloxacin plus rifampin appeared to be the most effective treatment for preventing invasive staphylococcal infections, although both ofloxacin and ofloxacin plus rifampin also showed antistreptococcal activity.

View larger version (18K): [in this window] [in a new window]   Figure 1. Mean number of infections in patients treated with norfloxacin, ofloxacin, or ofloxacin plus rifampin. Infections were caused by staphylococci (first set of diagonally striped bars; 0.447, 0.237, and 0.029 infections in patients treated with norfloxacin, ofloxacin, and ofloxacin plus rifampin, respectively), streptococci (black bars; 0.237, 0.026, and 0.029 infections), other microbiologically documented infections (second set of diagonally striped bars, 0.448, 0.158, and 0.085 infections), and clinically documented infections and unexplained fevers (white bars; 0.315, 0.737, and 0.886 infections).

The reduction in the rate of microbiologically documented infections in patients receiving ofloxacin or ofloxacin plus rifampin was offset by concomitant increase in unexplained fevers (fever occurred in 24% of norfloxacin recipients, 53% of ofloxacin recipients, and 49% of ofloxacin plus rifampin recipients; P = 0.02) (Table 2). The three treatment groups had a similar overall incidence of febrile neutropenic episodes as defined for this trial (79% of the norfloxacin group, 82% of the ofloxacin group, and 77% of the ofloxacin plus rifampin group; P > 0.2) (Table 2). Although the three treatment groups were similar in the time to the onset of the infectious episodes (median time to infection was 10.2 days for norfloxacin recipients, 10.4 days for ofloxacin recipients, and 12.6 days for ofloxacin plus rifampin recipients; P > 0.2), bacteremias occurred earlier in the norfloxacin group (P < 0.01) (Figure 2). The use of etoposide in the cytotoxic regimen was an independent predictor of bacteremic infection in the Cox regression model (P = 0.02). After adjustment for the effect of etoposide, only ofloxacin plus rifampin (relative to norfloxacin) influenced the time to bacteremia in the Cox regression model (P < 0.01). Treatment group did not influence the proportion of patients in whom the infectious episode had resolved at examination 3 days and 28 days after parenteral antimicrobial therapy (P > 0.2 for both times).

View larger version (17K): [in this window] [in a new window]   Figure 2. Kaplan-Meier product limit plot of the time to bacteremic infection (P < 0.01). Chart shows patients at risk for infection on given days by treatment group.

Discussion

Top Methods Results Discussion References

The results of this study show that oral quinolone-based antibacterial prophylaxis regimens with augmented gram-positive activity effectively suppress selected gram-positive infections in neutropenic patients with cancer. Ofloxacin, a fluoroquinolone that has enhanced in vitro gram-positive activity [10], had a greater protective effect against gram-positive infection than did norfloxacin, particularly against viridans streptococci. Rifampin, an inhibitor of DNA-directed RNA polymerase [28] that has an in vitro antimicrobial spectrum that includes coagulase-negative staphylococci and viridans streptococci [29], was highly effective in suppressing coagulase-negative staphylococcal infection when administered with ofloxacin. However, the combination appeared to be no more effective for streptococcal infection than was ofloxacin alone. The results were similar for intention-to-treat analyses that regarded the ineligible and inevaluable patients as having either treatment success or treatment failure (data not shown). Our finding that ofloxacin with or without rifampin was more effective than norfloxacin in reducing the rate of microbiologically documented infections was similar to the pattern reported in a large Italian study [30] comparing ciprofloxacin and norfloxacin.

Some limitations to our study may have influenced our results. For example, the open design required that patients assigned to the ofloxacin plus rifampin group receive two extra tablets daily, possibly influencing the incidence of gastrointestinal toxicity. Further, we used a protocol that was sufficiently restrictive to avoid biases that might have confounded the assessment of efficacy; however, we found that this curtailed enrollment. For example, the exclusion of patients with neutrophil counts less than 0.5 x 10⁹/L at study entry, fever at study entry, or age greater than 70 years accounted for almost half of all patients evaluated for study entry at one center. Although this selection bias may have somewhat reduced the generalizability of our results, it was probably responsible for allowing us to see a treatment effect that may have otherwise remained undetected.

The extent of the spectrum of gram-positive activity in antibacterial prophylaxis strategies for neutropenic patients has been studied using various agents. This approach has been considered important by clinicians who are concerned about the increasing problem of invasive gram-positive infection—particularly viridans streptococcal bacteremia—in patients with acute leukemia and bone marrow transplant recipients [31, 32]. The use of erythromycin in prophylaxis regimens has not been successful in reducing the risk for gram-positive infection [16-18]. Results with other agents, such as roxithromycin [19] and penicillin [15, 20, 21], have been more encouraging, although an increased incidence of penicillin-resistant streptococcal infections has been noted [21]. Prophylactic vancomycin has been successfully used to reduce the risk for gram-positive infections [33-35], but widespread use of this agent in some oncology and transplantation centers has been associated with the emergence of life-threatening vancomycin-resistant gram-positive infections [36]. Similar to our experience, oral administration of ciprofloxacin plus rifampin has been shown to have a protective effect against gram-positive bacteremia in a high-risk group of marrow autograft recipients who had prolonged neutropenia [34].

Gut epithelial damage induced by cytotoxic therapy has been implicated in the pathogenesis of invasive microbial disease, including viridans streptococcal bacteremia [21] and invasive fungal infection [37]. We saw a relation between the incidence of bacteremic infection and the use of etoposide-containing cytotoxic regimens, which are linked to increased mucosal toxicity in some patients with acute myeloid leukemia [37, 38]. The pathogenesis of these infections appears to involve the interaction of the damaged tissue with microorganisms that colonize the gut mucosal surfaces, followed by adherence, proliferation, translocation, and systemic infection [39].

Quinolone-based antibacterial chemoprophylaxis has successfully reduced the risk for infection caused by aerobic gram-negative bacilli from a range of 17% to 39% to a range of 0% to 11% in neutropenic patients having cytotoxic therapy for acute leukemia and bone marrow transplantation [1, 2, 4-740, 41]. We have confirmed these earlier observations. Only one (0.9%) aerobic gram-negative bacillary infection was seen, a urinary tract infection with norfloxacin-susceptible E. coli in a norfloxacin recipient. Aerobic gram-negative bacilli were cleared from rectal cultures in all patients after a median of 5.5 days. The median time to clearance suggested that norfloxacin and ofloxacin plus rifampin were more effective than ofloxacin alone. The increasing use of quinolones in community and hospital settings has been associated with the emergence of quinolone resistance among aerobic gram-negative bacilli [42-44], particularly among neutropenic patients with cancer receiving quinolone-based prophylaxis. This experience has paralleled that with trimethoprim-sulfamethoxazole prophylaxis [3, 45]. In general, infections in neutropenic patients with cancer that are caused by quinolone-resistant aerobic gram-negative bacilli have not been a problem in Canada or the United States [5, 7, 34, 41]. However, experience in Europe suggests that the continued prophylactic success of the quinolones against these pathogens may be confounded by the increasing prevalence of quinolone resistance [45]. To prevent the emergence of resistant gram-negative bacteria, maintenance of prophylactic efficacy may require additional agents, such as colistin as used with trimethoprim-sulfamethoxazole [4, 6, 46] or rifampin as used in this study and that of Gilbert and colleagues [34].

The development of rifampin resistance may limit the efficacy of this agent in prophylaxis regimens [29]. Further, animal data suggest that rifampin-quinolone combinations might promote the emergence of enterococcal infection [22]. Despite these theoretical considerations, enterococcal infection did not appear to be a problem in our study or in a study from Duke University [34].

There has been concern that prophylaxis based on oral quinolone may increase the risk for colonization and infection by opportunistic fungi [47]. We saw colonization by opportunistic yeasts in 71% of patients from one center. This finding was similar to the 76% reported in a study from the Johns Hopkins Oncology Center among recipients of norfloxacin antibacterial prophylaxis [47]. Oropharyngeal and rectal colonization occurred more frequently than nasopharyngeal colonization. We noted a trend toward a greater risk for colonization among ofloxacin plus rifampin recipients in this trial; however, like the experience from Duke University [34] among bone marrow autograft recipients receiving antibacterial prophylaxis with ciprofloxacin plus rifampin, the incidence of invasive fungal superinfections did not increase among ofloxacin plus rifampin recipients. The full antibacterial benefit of regimens, such as ofloxacin plus rifampin, may require the use of additional antifungal prophylaxis to offset the potential disadvantages of increased fungal colonization, particularly in the setting of gut epithelial damage induced by cytotoxic therapy.

Most antibacterial prophylaxis studies in neutropenic patients have not reduced the need for empiric, broad-spectrum parenteral antibiotic therapy, despite reductions in documented infections [3]. Our study was no exception. The hierarchical reduction in the rate of microbiologically documented infections was offset by a hierarchical increase in the rate of unexplained fevers (24% for norfloxacin recipients, 53% for ofloxacin recipients, and 49% for ofloxacin plus rifampin recipients; P = 0.02). Although unexplained fevers have been reported to result from increased absorption of such pyrogenic substances as endotoxin through gut epithelial surfaces damaged by cytotoxic therapy and not always from invasive infection [48], we could not differentiate between infectious and noninfectious pyrogenic processes as the causes of these unexplained fevers. A report from the Netherlands [49] showed that empiric parenteral antibacterial therapy could be safely withheld or discontinued early in neutropenic patients who are receiving quinolone prophylaxis and have unexplained fevers; this finding further supports the hypothesis that many such episodes do not represent infection. Similarly, a study from the University of Manitoba [7] showed that parenteral therapy for infection with gram-negative organisms could be safely discontinued after 24 to 36 hours in febrile neutropenic patients receiving oral quinolone. These reports lead us to speculate that our inability to discriminate infectious, unexplained fevers from those of noninfectious origin may have prevented us from recognizing the full protective potential of ofloxacin plus rifampin. This has important clinical and fiscal implications and should be addressed in future clinical trials.

Our results show that the benefit of antibacterial prophylaxis based on oral quinolone can be enhanced by selecting a quinolone with broader gram-positive activity and by adding agents with enhanced gram-positive activity, such as rifampin. However, the clinical significance of this benefit will not be fully understood until we better understand unexplained fevers.

Appendix

In addition to the National Cancer Institute of Canada Clinical Trials Group, the following investigators participated in this trial: R. Barr, MD, and A. Smith, MD (University Hospital, London, Ontario, Canada); E.J. Bow, MD (Health Sciences Centre, Winnipeg, Manitoba, Canada); R. Feld, MD (Princess Margaret Hospital, Toronto, Ontario, Canada); G. Garber, MD (Ottawa General Hospital, Ottawa, Ontario, Canada); D. Haase, MD (Victoria General Hospital, Halifax, Nova Scotia, Canada); R. Horn, MD, and H. Robson, MD (Royal Victoria Hospital, Montreal, Quebec, Canada); A. Keating, MD (The Toronto Hospital, Toronto, Ontario, Canada); W.S. Lofters, MD, and J. Pater, MD (Kingston General Hospital, Kingston, Ontario, Canada); T.J. Louie, MD (Calgary General Hospital, Calgary, Alberta, Canada); L.A. Mandell, MD (Henderson General Hospital, Hamilton, Ontario, Canada); P. Phillips, MD (Vancouver General Hospital and St. Paul's Hospital, Vancouver, British Columbia, Canada); R. Saguinur, MD (Ottawa Civic Hospital, Ottawa, Ontario, Canada); G. Taylor, MD (Walter MacKenzie Health Sciences Centre, Edmonton, Alberta, Canada); and K.E. Williams, MD (Royal University Hospital, Saskatoon, Saskatchewan, Canada).

Dr. Mandell: McMaster Medical Unit, Henderson General Hospital, Concession Street, Hamilton, Ontario L8V 1C3, Canada.

Dr. Louie: 841 Centre Avenue, NE, Calgary, Alberta T2E 0A1, Canada.

Dr. Feld: Princess Margaret Hospital, Room 597, 500 Sherbourne Street, Toronto, Ontario M4X 1K9, Canada.

Drs. Pater and Zee and Mr. Palmer: National Cancer Institute of Canada, Clinical Trials Group, Queen's University, 82-84 Barrie Street, Kingston, Ontario K7L 3N6, Canada.