Corresponding author and reprint requests: Alasdair P. MacGowan, Bristol Centre for Antimicrobial Research and Evaluation, Southmead Health Services NHS Trust and University of Bristol, Department of Medical Microbiology, Southmead Hospital, Westbury-on-Trym, Bristol BS10 5NB, UK, Tel: +44 117 9595654 Fax: +44 117 9593154 E-mail:
Objective: To study the interactions of colistin (MIC 2 mg/L) at concentrations of 0.5 and 5 mg/L with ceftazidime (1 and 75 mg/L, MIC 0.5 mg/L), aztreonam (1 and 30 mg/L, MIC 0.12 mg/L), meropenem (1 and 25 mg/L, MIC 0.03 mg/L), gentamicin (1 and 10 mg/L, MIC 2 mg/L), piperacillin (5 and 100 mg/L, MIC 4 mg/L) and ciprofloxacin (0.25 and 4 mg/L, MIC 1 mg/L) using a representative strain of Pseudomonas aeruginosa isolated from a cystic fibrosis patient.
Methods: The method used was a bacterial time kill curve with single agents and combinations. Using inocula of 106 CFU/mL, multiple sampling was performed over 6 h and in triplicate. The AUBKC of the time versus viable count curve, with single agents and combinations of agents, was taken as the endpoint for comparison.
Results: For colistin plus ceftazidime, colistin plus aztreonam, colistin plus meropenem and colistin plus ciprofloxacin, the pattern was for all the combinations (high or low concentrations) to produce smaller AUBKCs than single agents. In experiments using a bacteriostatic agent such as ceftazidime, the AUBKCs (log CFU/μL per h) for colistin 0.5 mg/L or 5 mg/L alone were 32.3±0.8 or 12.7±0.5, and for ceftazidime 1 mg/L or 75 mg/L alone they were 24.3±1.5 or 20.9±2.7. Combinations of colistin 0.5 mg/L plus either ceftazidime 1 mg/L or 75 mg/L produced AUBKCs of 23.8±1.8 or 16.1 mg/L. Combinations of colistin 5 mg/L plus ceftazidime 1 mg/L or 75 mg/L produced AUBKCs of 12.2±0.8 or 8.7±1.0. The AUBKCs for colistin 5 mg/L plus 75 mg/L are significantly smaller than those for the single agents, indicating synergy. In experiments using the bactericidal agent ciprofloxacin, the AUBKCs (log CFU/mL per h) for colistin 0.5 mg/L or 5 mg/L alone were 33.6±1.9 or 11.2±2.4, and for ciprofloxacin 0.25 mg/L or 4 mg/L alone they were 32.8±1.3 or 5.0±0.7. Combinations of colistin 0.5 mg/L plus either ciprofloxacin 0.25 mg/L or 4 mg/L produced AUBKCs of 32.2±0.9 or 4.3±1.4. Combinations of colistin 5 mg/L plus ciprofloxacin 0.25 mg/L or 4 mg/L produced AUBKCs of 10.7±1.5 or 4.2±0.6. Although combination AUBKCs were smaller than those for single agents, in no case did this reach statistical significance (p < 0.05).
Conclusions: These studies indicate that addition of colistin to other antipseudomonal drugs tends to produce smaller AUBKCs and hence greater killing of Pseudomonas aeruginosa than monotherapy.
Colistin has been available since the 1950s for clinical use in the treatment of Gram-negative infection . In the 1960s it was largely replaced in clinical practice, first by aminoglycosides and then later by β-lactams, with ever improving in vitro activity against Gram-negative bacteria. The last decade has, however, seen increasing resistance in Gram-negative bacteria, especially Pseudomonas aeruginosa and Acinetobacter spp.
Colistin retains excellent in vitro activity even for P. aeruginosa isolates from patients with cystic fibrosis  and is commonly administered by nebulizer to these patients, but also by the intravenous route [3,4]. In such clinical situations, patients are almost invariably receiving other antimicrobials, and the potential interactions between colistin and these agents have been poorly studied, most often using the checkerboard technique [5,6]. This method has its supporters  but has been criticized by others because of poor reproducibility and the use of arbitrary interpretative criteria [8,9]. We recently proposed another method of studying antimicrobial-antimicrobial interactions using bacterial time kill curves, with multiple sampling over 6–8 h followed by calculation of the area under the bacterial time killing curve (AUBKC) to compare the bactericidal activity of agents as monotherapies or in combination. In our test system we use pharmacokinetically achievable serum concentrations .
In this study we used bacterial time kill curves with both high and low pharmacokinetically achievable concentrations of colistin, plus ceftazidime, aztreonam, meropenem, piperacillin, gentamicin or ciprofloxacin, to study interactions assessed by the use of AUBKC as an endpoint, but also ‘classical’criteria .
Materials and Methods
A single strain of P. aeruginosa, recovered from a cystic fibrosis patient, was used in the study. The strain was stored at – 70°C until used.
The following antimicrobial agents were used: colistin sulfomethate sodium (Pharmax, Bexley, UK) at 0.5 and 5 mg/L; ciprofloxacin (Bayer, Newbury, UK) at 0.25 and 4 mg/L; gentamicin (David Bull Laboratories, Warwick, UK) at 1 and 10 mg/L; piperacillin (Lederle Laboratories, Maidenhead, UK) at 5 and 100 mg/L; ceftazidime (Glaxo Laboratories Ltd, Greenford, UK) at 1 and 75 mg/L; aztreonam (Bristol Myers Squibb Pharmaceutical Ltd, Dublin, Ireland) at 1 and 30 mg/L; and meropenem (Zeneca Pharma, Wilmslow, UK) at 1 and 25 mg/L.
Time kill curve method
The time kill curve methods were performed as described elsewhere . IsoSensitest broths (20 mL) (Unipath, Basingstoke, UK) were inoculated to give a final inoculum of 106 CFU/mL. For each antimicrobial, drugs were incorporated in nine broths, the agent alone in four and at the concentrations given above in four, plus a growth control. These were then sampled, and diluted as necessary, and viable counts were performed using a spiral plater (Spiral Systems, Don Whitley, Shipley, UK) at time 0, 45 min, and then every 20 min up to 3 h, and every 45 min up to 6 h, on nutrient agar plates (Difco, West Molesey, UK). Viable counts were read manually after incubation at 37°C. The study was performed in triplicate.
Time kill curves were drawn by plotting log10 CFU/mL against time (h), using the software package GraphPad Prism (GraphPad Software Inc., San Diego, USA). Data were log transformed and a baseline removed to compensate for the variability of the technique. AUBKCs were then calculated and compared using Student's t-test.
The results of a single experiment are shown in Figure 1. This shows the time kill kinetics of colistin alone and colistin plus ceftazidime, aztreonam. meropenem, piperacillin, gentamicin or ciprofloxacin. The MICs for the strain used were: colistin 2 mg/L: ceftazidime 0.5 mg/L; aztreonam 0.12 mg/L; meropenem 0.03 mg/L, piperacillin 4 mg/L; gentamicin 2 mg/L; and ciprofloxacin 1 mg/L.
Calculation of the AUBKCs showed that colistin alone was more bactericidal at a concentration of 5 mg/L (smaller AUBKC) than 0.5 mg/L. This was not true of ceftazidime, aztreonam, meropenem or piperacillin which were more or less equipotent at high or low concentrations (Table 1). Gentamicin and ciprofloxacin also had smaller AUBKCs at low concentrations than high. With all the six agents tested, the addition of colistin at a concentration of 5 mg/L increased the amount of killing, with low, high or both concentrations of the second drug. Ceftazidime and aztreonam increased killing in combination with colistin, and this was significantly greater than for either colistin or the second agent alone. Although not reaching statistical significance, there was a trend for other combinations to have smaller AUBKC values than single agents (Table 1).
Table 1. Areas under the bacterial time kill curves for colistin alone, second agents alone and combinations
Area under bacterial time kill curve (0–6 h) (log CFU/ml per h) (mean±standard deviation)
Colistin 0 mg/L
Colistin 0.5 mg/L
Colistin 5 mg/L
a Combination with colistin gave a significantly smaller AUC than second drug alone (p < 0.05).
b Combination gave a significantly smaller AUC than either colistin or second drug alone (p < 0.05).
According to ‘classical’interpretative criteria for synergy in time kill kinetic plots, i.e. the combination is >2 log more active at 24 h of incubation than the most active individual agent , then only colistin 5 mg/L plus ceftazidime 1 or 75 mg/L showed synergy in all three replicates. Colistin 5 mg/L also showed synergy with meropenem 1 mg/L or gentamicin 1 mg/L in two out of three replicates, and colistin 0.5 mg/L showed synergy in one of three replicates with ceftazidime 1 or 75 mg/L (data not shown).
These data are in broad agreement with others in showing that colistin, like aminoglycosides, quinolones and carbapenems, has bactericidal action against P. aeruginosa . However, it is less easy to compare our findings with previous data on antibiotic interactions with colistin. Checkerboard results have led to the suggestion that colistin plus gentamicin (using subinhibitory concentrations) is synergistic; synergy or addition was observed with colistin plus ceftazidime or impenem, and antagonism with colistin plus ciprofloxacin [5,6,13]. However, it is also known that comparison of checkerboard results with those from time kill curves results in less than 50% agreement, even when the experiments were done by the same workers . It is therefore unsurprising that we failed to confirm antagonism between colistin and ciprofloxacin. Indeed, our data based on AUBKC calculations showed greater killing by the combinations than by the individual agents. Combinations of colistin and ceftazidime or aztreonam demonstrated synergy according to statistical comparisons of AUBKC: that is, the combination was more active than the individual agents alone. This was supported by use of ‘classical’, though arbitrary, interpretative criteria for colistin plus ceftazidime but not for colistin plus aztreonam. Previously, we showed that the AUBKC was the most robust measure of the overall pattern of bacterial killing in a time kill kinetic plot . This parameter can also be used for statistical comparisons, and is thus preferable to the previously used criteria.
In conclusion, this work showed no evidence of antagonism between colistin and any of the agents tested. Addition of colistin at high concentration improves killing of P. aeruginosa compared to most of the second agents tested, especially at low concentrations, and synergy was observed with colistin plus ceftazidime or aztreonam.
We would like to thank Dr Lynda Fenelon, Department of Microbiology, St Vincents Hospital, Dublin for providing the strain of Pseudomonas aeruginosa used, and Dr Martin Goldman of Pharmax Ltd, Bexley, Kent, who provided advice and arranged the financial support.
The data in this manuscript were presented as a poster at the 37th Interscience Conference on Antimicrobial Agents and Chemotherapy, Toronto and published in abstract form.