Journal of Antimicrobial Chemotherapy (2003) 52, 668 674 DOI: 10.1093/jac/dkg403 Advance Access publication 1 September 2003 Treatment and outcome of Pseudomonas aeruginosa bacteraemia: an antibiotic pharmacodynamic analysis Sheryl A. Zelenitsky 1,2,5 *, Godfrey K. M. Harding 2,3,4, Siyao Sun 1, Kiran Ubhi 5 and Robert E. Ariano 1,2,5 Faculties of 1 Pharmacy and 2 Medicine, University of Manitoba; 3 Microbiology Laboratory, 4 Infectious Diseases, and 5 Pharmacy, St Boniface General Hospital, Winnipeg, MB, Canada Received 10 March 2003; returned 29 May 2003; revised 3 June 2003; accepted 6 July 2003 Objectives: To conduct a retrospective study of antibiotic pharmacodynamics in the treatment of Pseudomonas aeruginosa bacteraemia, and to identify pharmacodynamic indices associated with clinical cure. Methods: Cases of P. aeruginosa bacteraemia were identified, and information related to patient demographics, clinical status, antibiotic treatment and clinical outcome were documented. Anti-pseudomonal therapy was assessed, and concentration versus time profiles were constructed using measured levels for aminoglycosides, or population pharmacokinetic models for other antibiotics. P. aeruginosa isolates from all patients were retrieved and MICs for the anti-pseudomonal agents used to treat the episode of bacteraemia were determined. Patient- and treatment-related factors were tested for associations with clinical outcome using univariate and multivariate analyses. Results: Fifty cases of P. aeruginosa bacteraemia were identified and 38 cases were included in the pharmacodynamic analysis. Eighty-seven percent of patients received an aminoglycoside or ciprofloxacin and 79% received piperacillin or ceftazidime. A majority of patients, 71%, were administered a combination of antibiotics. Treatment outcomes were documented as persistent infection in 21%, death within 2 30 days in 21% and clinical cure in 58% of cases. Peak/MIC (P = 0.001) and AUC 24 /MIC (P = 0.002) for aminoglycosides and ciprofloxacin were significant factors in univariate tests. Only peak/mic was associated independently with treatment outcome (P = 0.017) in logistic regression analysis. The predicted probability of cure was 90% when peak/mic was at least 8. Conclusion: Pharmacodynamic considerations including aggressive dosing with targeted peak/mics for aminoglycosides and ciprofloxacin are strongly associated with clinical outcome and essential to the appropriate management of P. aeruginosa bacteraemia. Keywords: pseudomonal, bloodstream infections, peak/mic, AUC/MIC Introduction Pseudomonas aeruginosa bacteraemia is a serious and life-threatening infection especially in the immunocompromised and other susceptible populations. Reported mortality rates vary significantly from 20% 70% depending on patient- and infection-related factors. In previous investigations, variables associated with increased risk of death were pneumonia, 1 5 advanced age, 6,7 neutropenia, 2,4,8 serious underlying disease 4,5,7,9 and septic shock. 1,4 9 Predictably, the emergence of antibiotic resistance, 10 and use of inappropriate antibiotics have also been associated with death for patients with P. aeruginosa bacteraemia. 1,5,7,8 Studies by Bodey et al., 1 Chen et al. 7 and Kuikka & Valtonen 9 found increased mortality rates with aminoglycoside monotherapy. In another study, the risk of death from P. aeruginosa bacteraemia was 4.8 times higher for patients who did not receive at least one effective antibiotic in the treatment of their infection (P = 0.001). 8 Vidal and colleagues 5 also found increased mortality rates [odds ratio (OR) = 6.5, P = 0.04] in those who did not receive at least one agent active in vitro against P. aeruginosa administered in a dose and pattern considered appropriate by current medical standards. In more recent studies, similar associations between inappropriate antimicrobial treatment and mortality have been found for... *Correspondence address. Division of Clinical Sciences and Practice, Faculty of Pharmacy, University of Manitoba, Winnipeg, Manitoba, Canada R3T 2N2. Tel: +1-204-474-8414; Fax: +1-204-474-7617; E-mail: zelenits@ms.umanitoba.ca... 668 The British Society for Antimicrobial Chemotherapy 2003; all rights reserved.
Antibiotic pharmacodynamics in Pseudomonas aeruginosa bacteraemia critically ill patients with infections not limited to those caused by P. aeruginosa. 11 13 A strong association between inappropriate therapy and patient mortality has been established for P. aeruginosa bloodstream infections. Previous studies, however, confined their assessment of antibiotic therapy as appropriate or inappropriate based on in vitro susceptibilities and standard dosing regimens. A more descriptive and comprehensive approach can be provided by pharmacodynamic analyses that incorporate measures of antibiotic exposure (i.e. patientspecific antibiotic concentration profiles) and potency (i.e. MICs) to characterize the complex host pathogen antibiotic interaction. Associations can be made between indices such as peak/mic (i.e. peak concentration divided by MIC), T > MIC (i.e. percentage of the dosing interval with concentrations above MIC) or AUC 24 /MIC (i.e. area under the concentration time curve for 24 h divided by MIC) and microbiological or clinical outcome. The pharmacodynamic relationships can then be used to determine optimal antibiotic dosing strategies. This has been demonstrated in previous studies of ciprofloxacin for serious infections, 14 levofloxacin for respiratory tract, skin or urinary tract infections 15 and aminoglycosides for Gramnegative pneumonia. 16 Despite extensive study of the risk factors for poor outcome associated with P. aeruginosa infections, the impact of antibiotic pharmacodynamics on treatment response has not been investigated. The objectives of this retrospective study were to characterize antibiotic pharmacodynamics in the treatment of P. aeruginosa bacteraemia, and to identify pharmacodynamic indices associated with clinical cure. The goal was to provide clinically useful recommendations based on these results. Methods Patients and data collection This study was conducted at the St Boniface General Hospital, a 550 bed, tertiary-care facility. Approval for this study was obtained from the Health Research Ethics Board of the University of Manitoba. Cases of P. aeruginosa bacteraemia were identified from blood culture isolates entered into the Microbiology Laboratory database (Microscan, Dade Diagnostics Corp., Mississauga, Ontario, Canada) during 1996 2001, inclusive. Only cases of adult patients (i.e. 18 years) with accessible medical records were included. Hospital medical records were reviewed for information on patient demographics (i.e. age, gender, weight, height), serum creatinine, neutropenia (<1500 cells/mm 3 ), underlying diseases (i.e. chronic lung disease, ischaemic heart disease, congestive heart failure, renal or liver disease, diabetes mellitus, malignancy) and surgery or other invasive procedure within 30 days. Information related to the episode of P. aeruginosa bacteraemia including nosocomial acquisition (i.e. hospitalization >2 days), infection focus, clinical signs and symptoms and requirements for inotropes or mechanical ventilation were also documented. Anti-pseudomonal therapy initiated in response to a positive blood culture including antibiotics, doses and administration times was detailed. Standard practice included the use of manufacturer recommended dosing of antibiotics with appropriate adjustments for renal dysfunction. Aminoglycosides were administered as traditional or once-daily regimens, as selected by the attending medical team. Clinical outcome Clinical response to therapy was documented, and signs and symptoms of relapse were followed until discharge or for 30 days, whichever was less. Early mortality was defined as death within 2 days of a positive blood culture. Treatment failure was persistent bacteraemia or fever ( 38 C) exceeding 3 days, or death within 2 30 days of a positive blood culture. Clinical cure was resolution of all infection-related signs and symptoms without relapse. To evaluate the effects of antibiotic pharmacodynamics, cases with early mortality <2 days were excluded from the analysis. Pharmacokinetic/pharmacodynamic determinations Unbound, steady-state plasma concentration profiles for anti-pseudomonal agents administered during the first 3 days of treatment were simulated. For aminoglycosides, concentration versus time profiles were constructed using measured levels and a one-compartment pharmacokinetic model. Standard practice included the measurement of aminoglycoside peak and trough levels around the third dose. For ciprofloxacin, piperacillin and ceftazidime, profiles were predicted using population pharmacokinetic models and patient-specific data, including estimated creatinine clearance (Cl cr ) and body weight. Unbound, steady-state peak and trough concentrations were calculated according to: 17 C peak = f unbound Ro(1 e ket )/(Vd ke(1 e ket )) C trough = C peak (e ke(t t ) ) where C peak is the peak concentration at the end of the infusion, C trough is the trough concentration at the end of the dosing interval, f unbound is the fraction of drug not bound to plasma proteins, Ro is the infusion rate or dose divided by the infusion duration, ke is the elimination rate constant, t is the infusion duration, Vd is the volume of distribution and T is the dosing interval. Based on population pharmacokinetic data, estimates for piperacillin were t ½(normal) = 1 h, fe = 0.7, Vd = 0.28 L/kg and f unbound = 0.8, where t ½(normal) is the t ½ with normal renal function and fe is the fraction of drug excreted unchanged in the urine. The estimates for ceftazidime were t ½(normal) = 2 h, fe = 0.95, Vd = 0.3 L/kg and f unbound = 0.85, and those for ciprofloxacin were Vd = 2 L/kg and f unbound = 0.7. For piperacillin and ceftazidime, ke was determined from t ½, which was adjusted for renal function according to: 18 t ½(adj) = t ½(normal) /(1 fe(1 Cl cr /100)) and ke = t ½(adj) /0.693 Cl cr (ml/min) was estimated using: 19 Cl cr = (140 age) weight/scr 1.2 ( 0.85 for females) where scr is serum creatinine (µmol/l) that was measured during the episode of bacteraemia. For ciprofloxacin, ke was determined from total clearance (Cl T ) (L/h) according to: 20,21 Cl T = 0.173 Cl cr + 11.67 and ke = Cl T /Vd Dosing weights (DW) were used in pharmacokinetic calculations for patients with a body weight >130% of their ideal weight: 22,23 DW = ideal weight + 0.4(body weight ideal weight) for aminoglycosides and ciprofloxacin, and DW = ideal weight + 0.3(body weight ideal weight) for piperacillin and ceftazidime P. aeruginosa isolates from all patients were retrieved from the Microbiology Laboratory stock. MICs for the anti-pseudomonal agents used to treat the episode of bacteraemia were measured using macrodilution titres described by the NCCLS. 24 MICs were determined in triplicate on separate occasions. Pharmacodynamic indices were calculated using the simulated antibiotic concentrations and measured MICs. Peak/MIC was peak concentration at the end of the infusion 669
S. A. Zelenitsky et al. divided by the MIC, T > MIC was the percentage of the dosing interval with concentrations above the MIC and AUC 24 /MIC was the area under the concentration time curve for 24 h divided by the MIC. Peak/MIC for aminoglycosides and fluoroquinolones, T > MIC for β-lactams and AUC 24 /MIC for all classes of anti-pseudomonal agents were calculated. Total AUC 24 /MIC was also examined by adding the values for antibiotics used in combination. Statistical analysis Univariate and multivariate analyses were used to identify variables associated with early mortality within 2 days of a positive blood culture. Similarly, factors related to patient demographics, medical history, clinical status and antibiotic therapy were tested for associations with treatment outcome. Parametric data were presented as means and S.D., whereas non-parametric data were reported as medians and interquartile ranges. Students t-test, Mann Whitney U, Pearson χ 2 or Fisher s exact tests were used where appropriate. Significant variables from univariate analyses (α = 0.1) were included in a backwards stepwise logistic regression procedure to identify independent risk factors for treatment failure. P value limits of <0.1 and <0.05 were used for entrance into, and removal from, the model, respectively. Regression coefficients from the model were used to predict and graph the probability of clinical cure. Pharmacodynamic indices with significant associations were further assessed for breakpoints using classification and regression tree (CART) and receiver operating characteristic (ROC) curve analyses. All tests of significance were two-tailed. SPSS, Version 11 software was used for all statistical computations. Results Sixty-four cases of P. aeruginosa bacteraemia were identified in adult patients; however, 14 patients who were either not admitted to hospital or transferred to another institution were excluded. The characteristics of the 50 cases of P. aeruginosa bacteraemia studied are presented in Table 1. The average age was 70 ± 12 years for males and 66 ± 18 years for females. The male to female ratio was 1.2:1. Eighteen percent (9/50) of patients had neutropenia and 54% (27/50) had malignancies including 13 haematological cancers. Infections were nosocomial in 70% (35/50) of cases. Pneumonia and vascular catheters were the foci of infection in 36% (18/50) and 12% (6/50) of cases, respectively. Twenty-four percent (12/50) of patients died within 2 days of a positive blood culture. Based on multivariate analysis, female gender [P = 0.021, OR = 8.3, 95% confidence interval (CI): 1.4 49.7) and haematological malignancy (P = 0.006, OR = 11.7, 95% CI: 2.0 67.7) were independently associated with early mortality. The remaining 38 cases were included in the antibiotic pharmacodynamic analysis. Treatment failure was documented in 42% (16/38) of cases including eight patients who had persistent infection (i.e. three with persistent bacteraemia, five with fever >3 days) and eight patients who died. Clinical cure was achieved in 58% (22/38) of cases. Anti-pseudomonal therapy was initiated within 1 day (0.9 ± 0.2 days) of a positive blood culture. Aminoglycosides, ciprofloxacin, piperacillin or ceftazidime were administered in 53% (20/38), 42% (16/38), 42% (16/38) and 37% (14/38) of cases, respectively. Gentamicin was used in 18 cases and tobramycin in two cases. Antibiotic combinations were documented in 45% (17/38), 61% (23/38) and 71% (27/38) of cases on days 1, 2 and 3 of treatment, respectively. Combination regimens consisted of either an aminoglycoside or ciprofloxacin plus a β-lactam in 89% (24/27) of cases. Antibiotic Table 1. Patient demographic and infection characteristics for 50 cases of P. aeruginosa bacteraemia Characteristic pharmacokinetic parameters, MIC data and pharmacodynamic indices are summarized in Table 2. In univariate analyses (Table 3), peak/mic (P = 0.001) and AUC 24 /MIC (P = 0.002) for the aminoglycosides and ciprofloxacin were significantly associated with treatment response. Figure 1 shows the distribution of peak/mics and AUC 24 /MICs based on treatment outcome. Cure rates ranged from 27% (3/11) in patients with peak/ MICs < 4, to 90% (9/10) in those with ratios >6. Similarly, cure rates of 36% (4/11), 50% (6/12) and 90% (9/10) were associated with AUC 24 /MICs < 40, from 40 70 and > 70, respectively. In multivariate analysis, peak/mic was the only factor independently associated with treatment outcome (P = 0.017). The probability of clinical cure in relation to peak/mic is shown in Figure 2. The predicted probability of cure was 90% when peak/mic was at least 8. CART analysis identified two significant breakpoints, where cure rates were 0% (0/7) in patients with peak/mics < 2.9 and 84% (11/13) in those with values >4.8 (P = 0.001). ROC curve analysis identified a similar critical peak/mic of 4.6 for clinical cure (P = 0.001, sensitivity = 74%, specificity 71%). Discussion Mean ± S.D. or No. (%) patients Age (years) 68 ± 15 Gender (male) 27 (54%) Weight (kg) 74 ± 21 Neutropenia (<1500 cells/mm 3 ) 9 (18%) Chronic lung disease 9 (18%) Ischaemic heart disease 15 (30%) Congestive heart failure 10 (20%) Renal disease 11 (22%) dialysis 4 (8%) Diabetes mellitus 11 (22%) Malignancy 27 (54%) haematological 13 (25%) Procedure within 30 days 13 (26%) Nosocomial infection (>2 days) 35 (70%) Infection focus pneumonia 18 (36%) vascular catheter 6 (12%) urinary tract infection 6 (12%) other 9 (18%) unknown 11 (22%) White blood cell count (cells/mm 3 ) 9600 ± 8700 Inotropes 14 (28%) Mechanical ventilation 11 (22%) The overall mortality rate of 40% (20/50) in our study was in the range reported for P. aeruginosa bacteraemia. 6,9,25 Early mortality in 24% (12/50) of cases was also consistent with other investigations. 26 Female gender and haematological malignancy were independently associated with early mortality. Leibovici et al. 27 found higher mortality rates for hospital-acquired bacteraemia in women compared with men (43% versus 33%, P = 0.0001). Another study of 670
Antibiotic pharmacodynamics in Pseudomonas aeruginosa bacteraemia Table 2. Antibiotic pharmacokinetic parameters and pharmacodynamic indices in 38 patients with P. aeruginosa bacteraemia. Data presented as mean ± S.D. or median (interquartile range) ke (h 1 ) Vd (L) MIC (mg/l) Peak/MIC T > MIC (%) AUC 24 /MIC Gentamicin, tobramycin 0.14 ± 0.11 20.5 ± 11.4 2 (1 2) 3.2 (2.5 6.2) N/A 36 (29 65) (n = 20) Ciprofloxacin (n = 16) 0.14 ± 0.04 138.7 ± 31.6 0.5 (0.25 0.5) 4.7 (4.3 7.6) N/A 56 (50 72) Piperacillin (n = 16) 0.51 ± 0.09 20.0 ± 3.6 16 (8 32) N/A 59 (36 95) 41 (23 115) Ceftazidime (n = 14) 0.21 ± 0.13 20.3 ± 3.9 6 (4 32) N/A 100 (65 100) 146 (35 184) Table 3. Univariate analysis of risk factors for treatment failure in 38 patients with P. aeruginosa bacteraemia. Data presented as No.(%) patients, mean ± S.D. or median (interquartile range) Variable Clinical cure (n = 22) Clinical failure (n = 16) P value Age (years) 69 ± 14 64 ± 19 0.20 Gender (male) 15 (68.2%) 9 (56.3%) 0.45 Weight (kg) 77 ± 17 76 ± 27 0.53 Cl cr (ml/min) 57 ± 30 58 ± 30 1.0 Neutropenia (<1500 cells/mm 3 ) 3 (13.6%) 4 (25%) 0.68 Chronic lung disease 2 (9.1) 4 (25%) 0.22 Ischaemic heart disease 9 (40.9%) 5 (31.3%) 0.54 Congestive heart failure 5 (22.7%) 4 (25%) 1.0 Renal disease 5 (22.7%) 3 (18.8%) 1.0 Diabetes mellitus 4 (18.2%) 5 (31.3%) 0.45 Malignancy 10 (45.5%) 8 (50%) 0.78 Procedure within 30 days 7 (31.8%) 5 (31.3%) 0.97 Nosocomial infection (>2 days) 15 (68.2%) 11 (68.8%) 0.97 Pneumonia 5 (22.7%) 6 (37.5%) 0.32 Vascular catheter 3 (13.6%) 3 (18.8%) 0.68 Inotropes 5 (22.7%) 2 (12.5%) 0.68 Mechanical ventilation 3 (13.6%) 4 (25%) 0.43 Use of aminoglycoside 13 (59.1%) 7 (43.8%) 0.35 Use of ciprofloxacin 8 (36.4%) 8 (50%) 0.40 Use of piperacillin or ceftazidime 17 (77.3%) 13 (81.3%) 1.0 Use of combination antibiotics 16 (72.7%) 11 (68.8%) 0.79 Peak/MIC for aminoglycosides, 5.3 (4.5 9.5) 3.2 (2.1 4.6) 0.001 ciprofloxacin (n = 33) T > MIC (%) for piperacillin, 68 (31 100) 97 (47 100) 0.31 ceftazidime (n = 30) AUC 24 /MIC for aminoglycosides, 64 (49 94) 39 (25 51) 0.002 ciprofloxacin (n = 33) AUC 24 /MIC for piperacillin, 42 (20 158) 83 (31 181) 0.36 ceftazidime (n = 30) Total AUC 24 /MIC (n = 38) 94 (58 221) 80 (53 205) 0.73 892 patients admitted to surgical units showed that women were at higher risk of death from hospital-acquired pneumonia (OR = 2.25, P = 0.02) but not from other infections. 28 In our study, anti-pseudomonal therapy was initiated promptly within 1 day of a positive blood culture, and a combination of antibiotics was given to the majority of patients (i.e. 71%). There was no difference in treatment outcomes between patients who received monotherapy and those who were given a combination of antibiotics. Studies by Bodey et al. 1 and Chen et al. 7 found higher mortality rates in patients who received aminoglycoside monotherapy for P. aeruginosa bacteraemia. Hilf et al. 2 also concluded the superiority of combination over single antibiotics with mortality rates of 27% and 47%, respectively (P < 0.02). However, other investigations have not observed significant advantage with the use of multiple antibiotics. 25,29 Our study demonstrated the critical role of antibiotic pharmacodynamics in the treatment and outcome of P. aeruginosa bacteraemia. Where previous studies established the negative effects of using 671
S. A. Zelenitsky et al. Figure 1. Distribution of peak/mics and AUC 24 /MICs based on treatment outcome in 33 patients treated with gentamicin or ciprofloxacin for P. aeruginosa bacteraemia. (solid line represents medians). Figure 2. Probability of clinical cure in relation to peak/mic based on the logistic regression model (solid line). Data points for aminoglycosides (open triangles) and ciprofloxacin (open circles) were derived by forming ranges with 4 6 individual peak/mics and calculating mean probability of cure. Each case was then plotted using the mean probability of the respective range. inappropriate antibiotics based on susceptibilities and standard doses, ours detailed the influence of individualized dosing regimens and measured MICs. Although antibiotic treatments were all appropriate based on susceptible MIC breakpoints, pharmacodynamic profiles were considerably different due to inter-patient variability in pharmacokinetics and inter-isolate variability in MICs. In a relatively large study of 492 critically ill patients with bloodstream infections, inadequate antimicrobial treatment was an independent determinant of hospital mortality (OR = 6.86, P < 0.001). 12 However, the investigators noted high mortality rates for infections caused by some pathogens such as P. aeruginosa despite adequate treatment based on MICs. Although pathogen virulence and inflammatory response were proposed as explanations, our findings shift the paradigm towards one that implicates inadequate antibiotic pharmacodynamics. In our study, peak/mic was the only variable independently associated with treatment outcome. An important limitation and common criticism of retrospective studies is that such associations are susceptible to bias introduced by related, yet unidentified, variables. For example, could the increased risk of death be related to critical illness in patients with expanded volumes of drug distribution and thus lower peak concentrations? In other words, do lower peak/ MICs cause death or does another variable, such as critical illness, result independently in lower peak/mics and higher mortality rates? Firstly, our data were examined for variables associated with peak/mics less than or greater than the median value of 4. There were no differences in factors related to patient demographics, medical history and clinical status between those with low and high peak/mic values. Instead, lower peak/mics were explained by relatively reduced susceptibilities to both ciprofloxacin (median MICs of 0.5 versus 0.25 mg/l, P < 0.0001) and aminoglycosides (median MICs of 2 versus 1 mg/l, P = 0.07) and lower doses of aminoglycosides (medians 3.3 versus 4 mg/kg/day, P = 0.08). Secondly, concentration-dependent activity and the association between peak/mic and response have been well established for both aminoglycosides and fluoroquinolones. Retrospective studies published decades ago demonstrated the relationship between aminoglycoside peak concentrations and clinical outcome and established the use of therapeutic ranges. 30,31 Often, the indisputable evidence for pharmacodynamic relationships is provided by concentrationcontrolled animal studies. The required range to low concentrations with poor predicted response can be studied in animals without the impeding ethical considerations in humans. Drusano et al. 32 studied the pharmacodynamics of lomefloxacin, a fluoroquinolone, in the treatment of pseudomonal sepsis in neutropenic rats. Peak/MIC was associated with survival especially when higher values of 10 20 were obtained. The pharmacodynamic index, AUC/MIC, was also linked to outcome especially when correspondent peak/mics were <10. In a prospective, non-comparative trial, Preston and colleagues 15 analysed plasma concentration data and response in 134 patients who received levofloxacin for microbiologically confirmed infection. Peak/MICs 12 were associated with the best clinical and microbiological outcomes (P < 0.001). Kashuba et al. 16 studied aminoglycoside pharmacodynamics in the treatment of 78 cases of nosocomial Gram-negative pneumonia. Peak/MIC was the most significant index of clinical response as determined by temperature and leucocyte count resolution over 7 days. In patients with peak/ MICs 10, the predicted cure rate was 90%. These results were notably similar to those of our study. Although AUC 24 /MIC was significantly associated with clinical outcome in univariate tests, it was not an independent factor in multivariate analysis. The significant co-variance between peak/mic and AUC 24 /MIC makes the absolute separation of these parameters difficult. A determination of probability of cure versus AUC 24 /MIC using logistic regression analysis of our data predicted cure rates 90% when AUC 24 /MIC were at least 86 based on free concentrations, or 123 based on total concentrations (i.e. fraction unbound = 0.7). This threshold is almost identical to the significant breakpoint of 125 first 672
Antibiotic pharmacodynamics in Pseudomonas aeruginosa bacteraemia proposed by Forrest et al. in their study of ciprofloxacin in seriously ill patients. 14 Our pharmacodynamic analysis was strengthened considerably by the measured MICs and relatively high failure rate. Its limitations include the retrospective design, small sample size and estimated pharmacokinetic profiles for antibiotics other than the aminoglycosides. Nevertheless, we believe that aggressive aminoglycoside and ciprofloxacin dosing with targeted peak/mics of 8 10 is important in the successful treatment of P. aeruginosa bacteraemia. Of concern, however, is the difficulty in achieving these targets against pathogens such as P. aeruginosa with MICs that are relatively high but still reported susceptible. NCCLS breakpoints of 4 mg/l for aminoglycosides and 1 mg/l for ciprofloxacin can include a significant proportion of isolates for which pharmacodynamic profiles (i.e. peak/ MICs) would predict treatment failure. Within the current structure, clinicians must optimize aminoglycoside therapy with strategies such as once-daily dosing or using tobramycin instead of gentamicin if MICs are reduced. For ciprofloxacin, intensive dosing regimens of 400 mg or higher intravenously three times daily have been proposed. 14,33 In conclusion, peak/mic is an important determinant of treatment outcome for P. aeruginosa bacteraemia. Pharmacodynamic considerations are essential to the appropriate management of these serious infections. Acknowledgements This study was presented in September 2002 at the 42nd Interscience Conference on Antimicrobial Agents and Chemotherapy, San Diego, CA, USA. References 1. Bodey, G. P., Jadeja, L. & Elting, L. (1985). Pseudomonas bacteremia. Retrospective analysis of 410 episodes. Archives of Internal Medicine 145, 1621 9. 2. Hilf, M., Yu, V. L., Sharp, J. et al. (1989). Antibiotic therapy for Pseudomonas aeruginosa bacteremia: outcome correlations in a prospective study of 200 patients. American Journal of Medicine 87, 540 6. 3. Carratala, J., Roson, B., Fernandez-Sevilla, A. et al. (1998). Bacteremic pneumonia in neutropenic patients with cancer: causes, empirical antibiotic therapy, and outcome. Archives of Internal Medicine 158, 868 72. 4. Todeschini, G., Franchini, M., Tecchio, C. et al. (1998). Improved prognosis of Pseudomonas aeruginosa bacteremia in 127 consecutive neutropenic patients with hematologic malignancies. International Journal of Infectious Disease 3, 99 104. 5. Vidal, F., Mensa, J., Almela, M. et al. (1996). Epidemiology and outcome of Pseudomonas aeruginosa bacteremia, with special emphasis on the influence of antibiotic treatment. Analysis of 189 episodes. Archives of Internal Medicine 156, 2121 6. 6. Aliaga, L., Mediavilla, J. D. & Cobo, F. (2002). A clinical index predicting mortality with Pseudomonas aeruginosa bacteraemia. Journal of Medical Microbiology 51, 615 19. 7. Chen, S. C., Lawrence, R. H., Byth, K. et al. (1993). Pseudomonas aeruginosa bacteraemia. Is pancreatobiliary disease a risk factor? Medical Journal of Australia 159, 592 7. 8. Bisbe, J., Gatell, J. M., Puig, J. et al. (1988). Pseudomonas aeruginosa bacteremia: univariate and multivariate analyses of factors influencing the prognosis in 133 episodes. Reviews in Infectious Disease 10, 629 35. 9. Kuikka, A. & Valtonen, V. V. (1998). Factors associated with improved outcome of Pseudomonas aeruginosa bacteremia in a Finnish university hospital. European Journal of Clinical Microbiology and Infectious Diseases 17, 701 8. 10. Carmeli, Y., Troillet, N., Eliopoulos, G. M. et al. (1999). Emergence of antibiotic-resistant Pseudomonas aeruginosa: comparison of risks associated with different antipseudomonal agents. Antimicrobial Agents and Chemotherapy 43, 1379 82. 11. Harbarth, S., Ferriere, K., Hugonnet, S. et al. (2002). Epidemiology and prognostic determinants of bloodstream infections in surgical intensive care. Archives of Surgery 137, 1353 9. 12. Ibrahim, E. H., Sherman, G., Ward, S. et al. (2000). The influence of inadequate antimicrobial treatment of bloodstream infections on patient outcomes in the ICU setting. Chest 118, 146 55. 13. Kollef, M. H., Sherman, G., Ward, S. et al. (1999). Inadequate antimicrobial treatment of infections: a risk factor for hospital mortality among critically ill patients. Chest 115, 462 74. 14. Forrest, A., Nix, D. E., Ballow, C. H. et al. (1993). Pharmacodynamics of intravenous ciprofloxacin in seriously ill patients. Antimicrobial Agents and Chemotherapy 37, 1073 81. 15. Preston, S. L., Drusano, G. L., Berman, A. L. et al. (1998). Pharmacodynamics of levofloxacin: a new paradigm for early clinical trials. Journal of the American Medical Association 279, 125 9. 16. Kashuba, A. D., Nafziger, A. N., Drusano, G. L. et al. (1999). Optimizing aminoglycoside therapy for nosocomial pneumonia caused by gram-negative bacteria. Antimicrobial Agents and Chemotherapy 43, 623 9. 17. Sawchuk, R. J. & Zaske, D. E. (1976). Pharmacokinetics of dosing regimens which utilize multiple intravenous infusions: gentamicin in burn patients. Journal of Pharmacokinetics Biopharmaceutics 4, 183 95. 18. Tozer, T. N. (1974). Nomogram for modification of dosage regimens in patients with chronic renal function impairment. Journal of Pharmacokinetics and Biopharmaceutics 2, 13 28. 19. Cockcroft, D. W. & Gault, M. H. (1976). Prediction of creatinine clearance from serum creatinine. Nephron 16, 31 41. 20. Shah, A., Lettieri, J., Nix, D. et al. (1995). Pharmacokinetics of high-dose intravenous ciprofloxacin in young and elderly and in male and female subjects. Antimicrobial Agents and Chemotherapy 39, 1003 6. 21. Shah, A., Lettieri, J., Blum, R. et al. (1996). Pharmacokinetics of intravenous ciprofloxacin in normal and renally impaired subjects. Journal of Antimicrobial Chemotherapy 38, 103 16. 22. Wurtz, R., Itokazu, G. & Rodvold, K. (1997). Antimicrobial dosing in obese patients. Clinical Infectious Diseases 25, 112 18. 23. Bearden, D. T. & Rodvold, K. A. (2000). Dosage adjustments for antibacterials in obese patients: applying clinical pharmacokinetics. Clinical Pharmacokinetics 38, 415 26. 24. National Committee for Clinical Laboratory Standards. (2000). Methods for Dilution Antimicrobial Susceptibility Tests for Bacteria that Grow Aerobically: Approved Standard M7-A5. NCCLS, Wayne, PA, USA. 25. Siegman-Igra, Y., Ravona, R., Primerman, H. et al. (1998). Pseudomonas aeruginosa bacteremia: an analysis of 123 episodes, with particular emphasis on the effect of antibiotic therapy. International Journal of Infectious Diseases 2, 211 15. 26. Roche, O., Beuhorry-Sassus, F., Boillot, A. et al. (1995). [Pseudomonas aeruginosa septicemia. Host-related risk factors in 82 episodes]. Presse Medicale 24, 1164 66. 27. Leibovici, L., Paul, M., Weinberger, M. et al. (2001). Excess mortality in women with hospital-acquired bloodstream infection. American Journal of Medicine 111, 120 5. 28. Crabtree, T. D., Pelletier, S. J., Gleason, T. G. et al. (1999). Gender-dependent differences in outcome after the treatment of infection in hospitalized patients. Journal of the American Medical Association 282, 2143 8. 29. Chatzinikolaou, I., Abi-Said, D., Bodey, G. P. et al. (2000). Recent experience with Pseudomonas aeruginosa bacteremia in patients with cancer: retrospective analysis of 245 episodes. Archives of Internal Medicine 160, 501 9. 673
S. A. Zelenitsky et al. 30. Moore, R. D., Smith, C. R. & Lietman, P. S. (1984). The association of aminoglycoside plasma levels with mortality in patients with gram-negative bacteremia. Journal of Infectious Diseases 149, 443 8. 31. Moore, R. D., Lietman, P. S. & Smith, C. R. (1987). Clinical response to aminoglycoside therapy: importance of the ratio of peak concentration to minimal inhibitory concentration. Journal of Infectious Diseases 155, 93 9. 32. Drusano, G. L., Johnson, D. E., Rosen, M. et al. (1993). Pharmacodynamics of a fluoroquinolone antimicrobial agent in a neutropenic rat model of Pseudomonas sepsis. Antimicrobial Agents and Chemotherapy 37, 483 90. 33. Dan, M., Poch, F., Quassem, C. et al. (1994). Comparative serum bactericidal activities of three doses of ciprofloxacin administered intravenously. Antimicrobial Agents and Chemotherapy 38, 837 41. 674