Colistin pharmacokinetics: the fog is lifting


W. Couet, INSERM ERI 23, Pôle Biologie Santé (PBS), Médecine-Sud, Niveau 1, 1 rue Georges Bonnet, 86022 Poitiers Cedex, France


Clin Microbiol Infect 2012; 18: 30–39


Colistin is a re-emerging old antibiotic that is used to treat multidrug-resistant infections in critically ill patients. It corresponds to a mixture of at least 30 different compounds administered as inactive derivatives. Therefore, colistin pharmacokinetics are quite difficult to investigate and complex to predict. However specific chromatographic methods have been made available in recent years, leading to a series of modern pharmacokinetic studies after intravenous administration of the prodrug to critical-care patients; these have been conducted by a few groups and have only been recently published. The objective of this article was to conduct a critical review of these very informative modern pharmacokinetic studies and to provide prospective thoughts.


After being abandoned in the early 1980s because of reported nephrotoxicity and neurotoxicity, colistin is having a second life as a ‘salvage’ treatment against multiresistant Gram-negative bacteria such as Pseudomonas and Acinetobacter in critically ill patients [1,2]. Colistin is composed of at least 30 different polymyxin compounds, mainly colistin A and B, and it is administered intravenously as a complex mix of colistin methanesulphonate (CMS) derivatives. Initial pharmacokinetic studies relied on bioassays, which are still used in some institutions. However, because colistin is administered as a prodrug that can be easily converted into the active compound during storage or sample handling, accurate pharmacokinetic studies require specific analytical assays, and bioassays should therefore no longer be used [1]. Chromatographic procedures using HPLC with post-derivation fluorimetric detection [3] or liquid chromatography–tandem mass spectrometry (LC-MS/MS) [4,5] have only relatively recently been developed, and used by several groups in order to complete a series of major modern pharmacokinetic studies.

Several extensive reviews have been published on colistin in general and on its pharmacokinetics in particular [1,6,7]. The objective of the present article was not simply to update the previous reports, but also to conduct a critical review with prospective thoughts and a focus on the most recent pharmacokinetic studies, published during the last few years or months, or not yet published but only presented at international meetings. Furthermore, although CMS may be aerosolized, especially in patients with cystic fibrosis, this last mode of administration raises a number of specific biopharmaceutical questions that have been recently characterized in rats [8]. It will not be considered throughout this review, which will be focused on colistin pharmacokinetics in critical-care patients treated intravenously with CMS.

Pharmacokinetic Studies in Rats

An initial pharmacokinetic study in rats, using a novel specific HPLC assay, was conducted in 2003, after direct intravenous administration of colistin, allowing the estimation of primary pharmacokinetic parameters [9]. The colistin volume of distribution was estimated to 0.50 ± 0.06 L/kg, indicating limited extravascular distribution, consistent with its physicochemical characteristics, including molecular mass and hydrophilicity. Protein binding, which has an effect on distribution and elimination, was determined by equilibrium dialysis, as non-specific adsorption on the membranes precluded the use of ultrafiltration. The colistin unbound fraction (fu) varied between 43% and 45% as concentration ranged from 1.5 mg/mL to 6.0 mg/mL. However, colistin A was more extensively bound (fu ∼36%) than colistin B (fu ∼52%). Colistin clearance was estimated to be 5.2 ± 0.4 mL/min/kg, with <1% of colistin being excreted unchanged in urine. The derived clearance value (CLR = 0.010 ± 0.008 mL/min/kg) was much lower than the expected renal clearance by glomerular filtration, which, assuming an average fu of 0.44 and a glomerular filtration rate (GFR) of 5.2 mL/min/kg in rats, should be equal to 2.3 mL/min/kg. This low renal clearance suggested extensive tubular reabsorption, which was later confirmed by the same group and attributed to active transport systems [10]. As another consequence of this very low renal clearance, one may conclude that colistin non-renal clearance is almost equal to its total clearance and much lower than hepatic blood flow in rats (72–95 mL/min/kg), suggesting a low hepatic extraction ratio. Therefore, colistin clearance should be independent of hepatic blood flow but should vary with protein binding and hepatic intrinsic clearance. However, as colistin is probably not extensively bound to plasma proteins, its hepatic intrinsic clearance and therefore enzymatic activity could mainly determine its clearance value. Finally, the colistin elimination half-life in rats was 74.6 ± 13.2 min.

One year later, the same group conducted a second pharmacokinetic study in rats, but after intravenous administration of CMS at a dose of 15 mg/kg [11]. The same specific HPLC assay [3] was used, allowing direct comparisons between the two studies. However, CMS concentrations were not measured directly, but were obtained from the difference between colistin levels determined in samples after and before hydrolysis. Therefore, the authors mentioned that pharmacokinetic parameters for CMS may best be considered as hybrid parameters for CMS and the partially sulphomethylated derivatives. In that study, the CMS volume of distribution was estimated to be 0.30 ± 0.06 L/kg, corresponding almost exactly to extracellular fluid volume, and to about 60% of the value determined for colistin. CMS protein binding was not estimated, as ultrafiltration is not recommended, for reasons previously explained, and equilibrium dialysis cannot be used, because of unavoidable spontaneous hydrolysis during the process. CMS elimination was characterized by a total clearance of 11.7 ± 1.8 mL/min/kg. The fraction of the CMS dose recovered in urine was estimated to be 61% ± 14%, and 33% ± 15% was recovered as colistin. However, given the low urinary recovery of colistin (<1%) in rats treated directly with this active moiety [9], the authors concluded that most of the colistin measured in urine after CMS administration was the result of post-excretion CMS hydrolysis. Accordingly, CMS renal clearance was estimated with the summed amount of CMS and colistin recovered in urine. CMS renal clearance (7.2 ± 2.2 mL/min/kg) was only slightly greater than GRF in rats (5.2 mL/min/kg), and this difference could be within the limits of experimental uncertainty. Therefore, except in cases of extensive plasma protein binding, CMS would undergo modest net tubular secretion, as suggested by the authors. CMS non-renal clearance was estimated to be 4.9 mL/min/kg on average, and, assuming that colistin clearance was the same after this CMS administration and after direct injection [9], it was possible to estimate that only 6.8% of the CMS dose was converted into colistin on average. Finally, the CMS elimination half-life (23.6 ± 3.9 min) was significantly shorter than that of colistin (55.7 ± 19.3 min), indicating that colistin disposition is not rate-limited by its formation. However, colistin concentrations reached a peak very early (after 5 min in three rats and after 10 min in the remaining two), which is usually a characteristic of metabolites with formation rate-limited elimination [12]. Therefore, this early peak occurrence for a ‘metabolite’ with an elimination that is not rate-limited by its formation is unusual and difficult to explain.

More recently, a dose-ranging pharmacokinetic study was conducted by our group, using a new LC-MS/MS assay [13]. No trend of non-linearity was observed in this large range of doses (5–120 mg/kg), and systemic pharmacokinetic parameter values were fully consistent with those previously published by Li et al., in particular when the same dose (15 mg/kg) was administered. Using the same calculation and hypothesis as Li et al., we estimated that 15.5% of the CMS dose was converted into colistin in rats on average, which is slightly more than the value reported by these authors (6.8%), but low enough to confirm that, at least in rats, only a limited fraction of the CMS dose is actually converted into colistin. Interestingly, we also observed the previously described unexpected and still unexplained early peak appearance for a ‘metabolite’ with elimination that is not rate-limited by its formation. Considering that rapid hydrolysis of CMS in rat plasma could possibly explain this rapid colistin appearance, we compared the relative instability of CMS in spiked human vs. rat plasma at 37°C. However, no major difference was observed between the CMS rate of degradation and the colistin rate of formation in rat and human plasma (Fig. 1).

Figure 1.

 Colistin methanesulphonate (CMS) disappearance vs. time at 37°C in fresh human (□) and rat (bsl00001) plasma, spiked with CMS at an initial concentration close to 7.5 mg/mL, and corresponding colistin appearance in human (○) and rat (•) plasma.

The effect of hepatic insufficiency on colistin pharmacokinetics in rats was recently investigated in our laboratory (unpublished data, available on request). Acute hepatic failure was obtained by intraperitoneal administration of CCl4 and confirmed by alanine aminotransferase and aspartate aminotransferase determinations. Control rats and rats with hepatic failure (n = 6 per group) were treated 24 h after CCl4 administration, either with CMS (15 mg/kg administered intravenously) or with colistin (1.5 mg/kg administered subcutaneously). No significant differences were observed between CMS and colistin pharmacokinetic parameters estimated in rats with hepatic failure and control animals after CMS or colistin administration (Fig. 2). However, the possibility cannot be excluded that this absence of effect was attributable to the inability of the experimental model to reproduce and predict what would happen in patients with hepatic insufficiency.

Figure 2.

 Colistin methanesulphonate (CMS) (•) and colistin (○) plasma concentrations (mean ± SD) vs. time following single intravenous dose (15 mg/kg) of CMS in: (a) control rats; and (b) rats with impaired hepatic function.

Modern Pharmacokinetic Studies in Humans After intravenous Administration of CMS

A pharmacokinetic study of colistin following intravenous CMS administration to patients with cystic fibrosis and using a specific assay was published in 2003 [14]. However, as previously mentioned, CMS is most frequently aerosolized for cystic fibrosis patients, and this review is essentially focused on critical-care patients treated intravenously with CMS. For homogeneity purpose, CMS doses will be expressed in millions of International Units (MIU), which may not be the most appropriate practice [15], but is at least the most usual.

An initial pharmacokinetic study of colistin in critically ill patients, using a chromatographic assay, was published in 2008 by Markou et al. [16]. Fourteen patients were studied after at least 2 days of CMS treatment at a maintenance dose of 3 MIU three times daily. Noticeably, colistin concentrations were not determined after the initial CMS dose, and CMS concentrations were not assayed. Therefore, the major reported finding was a colistin plasma peak concentration of 2.93 mg/mL, occurring on average 15 min after the end of the 30-min infusion. However, such an early peak was not observed in the most recent studies, and it can most likely be explained, at least in part, by uncontrolled post-sampling CMS degradation, as only a low percentage of CMS hydrolysis would have a pronounced influence on early colistin concentrations.

Therefore, in our opinion, the really first major modern pharmacokinetic study in critical-care patients was published in 2009 by Plachouras et al. [17]. CMS was administered at a dose of 3 MIU (equivalent to 240 mg of CMS) infused over 15 min every 8 h, corresponding to a very common dose, and pharmacokinetic analysis was conducted after the first and fourth infusions (typically 2 × 9 samples per patient). CMS and colistin concentrations were measured with a novel LC-MS/MS assay [4], and a population pharmacokinetic analysis was conducted. Critically ill patients (n = 18, six females and 12 males; mean age of 63.6 years, with creatinine clearance (CrCL) ranging between 41 and 126 mL/min, with a mean value of 82.3 mL/min) were enrolled. CMS and colistin data were analysed simultaneously with a two-compartment model for CMS and a one-compartment model for colistin. The pharmacokinetic parameter values obtained in this study are presented in Table 1. The colistin elimination half-life (typical value: 14.4 h) was much longer than that of CMS (typical value: 2.3 h), indicating that colistin elimination is not formation rate-limited, as already observed in rats [11]. However, as opposed to the findings of Markou et al. [16], it took about 7 h on average before the colistin plasma peak concentration was reached, consistent with the fact that colistin distribution is rate-limited by its elimination, but different from the almost instantaneous colistin peak appearance in rats. One of the most interesting characteristics of this study was that data were collected for the first time after patients received the initial CMS dose. It clearly showed that, for a typical patient, colistin concentrations should stay below the MIC breakpoints (2 mg/mL) published by the European Committee on Antimicrobial Susceptibility Testing [18] after the first few 3 MIU/8 h doses, corresponding to what is probably the most common dosing regimen. This observation was extremely important, as insufficient antibiotic concentrations are likely to be responsible for increased mortality and/or resistance development. Therefore, the authors conducted simulations, suggesting the use of a loading dose of up to 9 MIU, and then various dosing intervals between two consecutive doses at steady state. The results of these simulations were later confirmed by another study conducted by the same group in ten critical-care patients who received a loading dose of 6 MIU (Friberg et al., 50th ICAAC, 2010, A1-665). Colistin plasma concentrations above 1 mg/mL were obtained within 4 h in the majority of patients. However, independently of the loading dose, colistin plasma concentrations at steady state are determined by the maintenance dose, and for a typical ‘Plachouras et al. patient’, the usual 3 MIU/8 h dosing regimen would most often lead to average colistin plasma concentrations at steady state lower than 2 mg/mL, which may not be enough for bacteria with reduced susceptibility to antibiotics. Another interesting finding resulting from the slow increase in colistin plasma concentrations with time and from the short dosing interval (8 h) as compared with the colistin half-life (14.4 h) was that the fluctuations in colistin plasma concentrations at steady state are limited. The authors did not observe a correlation between CrCL and colistin disposition, but acknowledged that this may be attributable to the limited number of patients and relatively small range of CrCL values, with an average of 82.3 mL/min.

Table 1.   Previously published [17,19,22] (Grégoire et al., 21st ECCMID, 2011, 804pp.) pharmacokinetic parameter values (relative standard error) for colistin methanesulphonate (CMS) and colistin after intravenous administration of CMS to healthy volunteers or critical-care patients
Healthy volunteerCritically ill patients
Couet et al. (22)Plachouras et al. [17]Garonzik et al. [19]Gregoire et al. (Grégoire et al., 21st ECCMID, 2011, 804pp.)
Typical valueIIV (CV%)Typical valueIIV (CV%)Typical valueIIV (CV%)Typical valueIIV (CV%)
  1. CL, clearance of CMS; CL/fm, clearance of formed colistin; CLR, renal clearance of CMS; CLR/fm, renal clearance of formed colistin; CV, coefficient of variation; IIV, inter-individual variability; Q, intercompartmental clearance for CMS; Vc, volume of distribution of central compartment for CMS; V/fm, volume of distribution of formed colistin; Vp, volume of distribution of peripheral compartment for CMS.

  2. aParameter value calculated from the authors’ formula for a creatinine clearance of 120 mL/min/1.73 m2.

  3. bCL/fcoli and V/fcoli are reported instead of CL/fm and V/fm.

 CL (mL/min)148 (5)15 (47)228 (10)37 (15)61.0 (–)
107 (–)42 (30)
 Vc (L)8.92 (6) 13.5 (45) 11.5 (5)325.3 (10) 
 Q (mL/min)41.4 (15) 2217 (35) 133 (12)84123 (13) 
 Vp (L)5.1 (–) 28.9 (22) 18.7 (9)7929.7 (12) 
 CLR (mL/min)103 (8)16 (48) 29.3 (9)
82.8 (10) 
 CL/fm (mL/min)48.7 (15) 151.5b (19)59 (36)45.3 (–)
2333.4 (–) 
 V/fm (L)12.4 (15)19 (53)189b (12) 45.1 (6)487.2 (11) 
 CLR/fm (mL/min)1.9 (19)56 (49)  7.0
 2.1 (17)63 (50)

The second major modern pharmacokinetic study of colistin after intravenous administration of CMS to critical-care patients was conducted on a larger scale (n = 105 patients for 851 samples analysed), and has just been published [19]. CMS was administered as a short-term infusion every 8–24 h. The median daily dose of colistin was 6.7 MIU (range: 2.5–13.6 MIU). Among the 105 patients, 89 were not on renal replacement, 12 were receiving haemodialysis, three were on continuous venovenous haemodialysis, and one was on continuous venovenous haemofiltration. CrCL values ranged from 0 to 169 mL/min/1.73 m2. A number of observations were similar to the findings of Plachouras et al. [17]. In particular, the distribution of CMS and colistin was best described by a linear model comprising, respectively, two compartments and one compartment. Colistin plasma concentration vs. time profiles were much flatter that those of CMS, consistent with the longer colistin elimination half-life. Average colistin steady-state plasma concentrations (Css,avg) varied in a relatively large range (0.48–9.38 mg/mL, with a median of 2.36 mg/mL), reflecting the greater variability in patient status. Pharmacokinetic parameter values are presented in Table 1. A strong inverse trend was observed between Css,avg and CrCL. Accordingly, CrCL was an important covariate, not only for CMS but also for colistin. The authors confirmed that the reason why CrCL was not identified as a covariate for colistin clearance in the Plachouras et al. study was simply the narrow range of CrCL. Furthermore, data from both studies are in agreement in suggesting that, in patients with normal or moderately impaired renal function, the currently recommended daily doses of CMS are not sufficient to obtain reliably efficacious steady-state concentrations of colistin. The authors proposed the following dosing algorithm, incorporating renal function, to estimate the CMS maintenance dose (expressed in colistin base activity) required to reach a target Css,avg (mg/mL) for colistin:

image(Eqn 1)

This daily dose may be expressed in CMS MIU as follows:

image(Eqn 2)

In these equations, CrCL is expressed in mL/min/1.73 m2. It was confirmed that CMS and colistin are efficiently cleared by venovenous haemodiafiltration [20] or during haemodialysis [21], and the algorithm was successfully adjusted to this situation. Overall CrCL explained approximately 60% of the variability in the ideal maintenance dose corresponding to a target Css,avg, and the authors recommended that CMS daily doses higher than 10 MIU should not be used.

The usefulness of starting CMS treatment with a loading dose, as suggested by Plachouras et al. as a consequence of the slow increase in colistin concentrations with time, was also considered. Because the central compartment volume for CMS was successfully modelled as a function of body weight, the authors recommended a CMS loading dose adjusted to body weight (using the lower of either actual or ideal body weight), followed by a maintenance dose starting 24 h later. Accordingly, a patient with a body weight of 60 kg would require 8 MIU of CMS to reach a target concentration equal to 2 mg/mL, but the authors recommended limiting the loading dose to a maximum of 10 MIU.

The first modern pharmacokinetic study of colistin following single-dose intravenous administration of CMS (1 MIU) to healthy volunteers was conducted by our group [22] and published almost simultaneously with the article of Garonzik et al. [19]. The slow appearance of colistin in plasma was again observed, and pharmacokinetic parameter values in healthy volunteers (Table 1) were essentially consistent with those previously obtained by Plachouras et al. [17] and Garonzik et al. [19] in critical-care patients. However, for the first time, urine samples were collected and analysed for CMS and colistin concentrations, in order to differentiate between renal and non-renal CMS clearance. The data were corrected for post-excretion hydrolysis of CMS to colistin. Accordingly, the fraction of the CMS dose recovered after 24 h in urine was equal to 49%, but the fraction actually excreted (fe) was estimated to be 64%. The typical value for CMS renal clearance was 103 mL/min, whereas that for colistin renal clearance was much lower (typical value of 1.9 mL/min), consistent with the extensive tubular reabsorption demonstrated in rats. On the assumption that the fraction of the CMS dose not excreted unchanged (1 – fe) was totally converted into colistin, as initially suggested by Li et al. [1], it was possible to determine that, in healthy volunteers, about 30% of the CMS dose was converted into colistin. The ‘true’ colistin pharmacokinetic parameters values could then be estimated, and not only the apparent pharmacokinetic parameters (parameters divided by the fraction of the administered CMS dose eventually converted into colistin), in order to facilitate comparisons with physiological parameter values such as extracellular body water for volume terms. However, Garonzik et al. [19] suggested that only a fraction of the CMS dose that was not excreted unchanged in urine was eventually converted into colistin. This is probably correct, because some of the partially sulphomethylated derivatives are likely to be excreted in urine before being converted into colistin. How much of the CMS dose not excreted unchanged in urine is not converted into colistin is difficult to assess. However, it is important to mention that these two articles, published almost simultaneously, used the same terminology (fm) with different meanings. In our article, fm refers to the fraction of the CMS dose that was not excreted unchanged in urine [22], whereas in Garonzik et al. [19], fm corresponds to the unknown fraction of the non-renal clearance of CMS that actually forms colistin, which we implicitly assumed to be equal to one. We therefore suggest keeping the terminology used by our colleagues, and not using fm as we did, but instead using 1 – fe, and we propose a new schematic representation of CMS and colistin distribution (Fig. 3) to summarize this terminology issue. Note that, with this terminology, the fraction of the CMS dose eventually converted into colistin is not equal to fm, and it was named fmcoli in order to avoid further confusion, with fmcoli = (1 − fe) × fm.

Figure 3.

 Schematic representation of the disposition of colistin methanesulphonate (CMS) and the colistin generated from it in the body following intravenous administration of CMS, adapted from the initial pathway proposed by Li et al. to account for potential incomplete conversion of partially sulphomethylated derivatives to colistin, possibly owing to urinary excretion of these intermediate compounds, where fe corresponds to the fraction of the CMS dose excreted unchanged in urine, and fm to the fraction of non-renally cleared CMS that actually forms colistin.

The last modern pharmacokinetic study to our knowledge was conducted by our group, and combined data collected in healthy volunteers with those obtained in critical-care patients (n = 32 for 242 samples analysed) treated in a real clinical setting. The results have not yet been published, but were presented at the last ECCMID meeting in Milan (Grégoire et al., 21st ECCMID, 2011, 804pp.). Patients had various levels of renal function, from normal to severely impaired, but none required extrarenal replacement. The dosing regimen varied between 0.5 and 4 MIU/8 h, with a median of 2 MIU/8 h. Colistin concentrations ranged from below the limit of quantification to 6.4 mg/mL, with a median of 1.1 mg/mL. Pharmacokinetic parameters were estimated with a population approach, using the previously reported pharmacokinetic model with two compartments for CMS and one compartment for colistin [17,19]. The pharmacokinetic parameters estimated in critically ill patients are presented in Table 1, and are consistent with those presented in the same table and obtained by others during these recent modern pharmacokinetic studies. It was confirmed that CMS renal clearance was related to CrCL, and, interestingly, the estimated colistin clearance in this study (CLcoli = 33.4 mL/min) had a tendency to be lower than previously reported values, in particular in healthy volunteers (typical value for CLcoli of 48.7 mL/min).

Altogether, these recent modern pharmacokinetic studies have provided a large and important amount of useful information. In particular, Garonzik et al. showed that using CrCL as a covariate explained about 60% of the variability in the ‘ideal’ maintenance dose of CMS required to achieve a given target Css,avg. However, 40%, which is about half of the variability, is still out of control. As a consequence, although a majority of patients exhibit colistin plasma concentrations at steady state within the target range, some individuals occasionally present unexpectedly high colistin concentrations, as observed by Garonzik et al. as well as by ourselves in our current practice. Attention should now be paid to these patients, as high colistin plasma concentrations may be responsible for toxicity. This question will now be investigated from a theoretical standpoint, in order to identify potential factors responsible for unusually high colistin concentrations.

Because colistin concentration fluctuations within a dosing interval at steady state are limited when CMS is administered three times daily, as is usually recommended, at least for the moment, it is essential to understand which pharmacokinetic parameters determine Css,avg and how these parameters may be altered in critical-care patients.

By definition, at steady state, the rate of colistin formation is equal to its rate of elimination, from which Eqn 3 can be derived.

image((Eqn 3))

where fmcoli corresponds to the fraction of the CMS dose converted into colistin, dose/τ the CMS maintenance dose, and CLcoli the colistin clearance. As this has no impact on the general reasoning but simplifies the demonstration, it will be assumed that the fraction of the CMS dose not excreted unchanged in urine is totally converted into colistin (fm = 1). Eqn 3 may then be reorganized, leading to Eqn 4, which predicts Css,avg.

image(Eqn 4)

Therefore, Css,avg is determined by two parameters reflecting formation (CMS clearance terms) and a third one characterizing elimination (colistin clearance). Notably, the colistin volume of distribution, which is likely to be altered in critical-care patients, has no effect on Css,avg. Simulations were conducted with parameter values estimated in healthy volunteers as a starting point [22]. CMS renal clearance, referred to as CLR, was assumed to be equal to 85% of GFR (CLR = 0.85 GRF), and CMS non-renal clearance (CLNR) and colistin clearance were set at 50 mL/min.

Effect of altered CMS renal clearance on Css,avg

Simulations were conducted for various maintenance doses (1, 2 and 3 MIU every 8 h), and Css,avg was predicted for various CrCL values as an indicator of changing CMS renal clearance (Fig. 4). These simulations show that, for an individual with normal renal function, the usual upper CMS dosing regimen (3 MIU/8 h) would lead to a Css,avg close to 2 mg/mL. It is also apparent from these simulations that Css,avg could not be higher than approximately 10 mg/mL, even in an anuric patient (CrCL = 0 mL/min). However, these simulations also suggest that a relatively high Css,avg could be reached in patients with altered renal function, provided that the CMS dosing regimen was maintained at a sufficient level. As an example, a Css,avg of 4 mg/mL could be reached in a patient with CrCL equal to 30 mL/min, provided that the 3 MIU/8 h CMS dosing regimen was maintained. Therefore, according to these simulations, moderate renal impairment should not be viewed only as causing a risk of toxicity, owing to overexposure, but also as an opportunity to reach efficient colistin concentrations. Reducing CMS doses in patients with renal insufficiency according to the recommendations appearing in product information before the Garonzik et al. article may not be appropriate. However, of course, clinical evidence, including careful monitoring of toxicity, should support this statement.

Figure 4.

 Simulated colistin concentrations at steady state (Css,avg) as a function of creatinine clearance (CrCL) following various maintenance doses of colistin methanesulphonate (CMS) (1 MIU/8 h (green), 2 MIU/8 h (fuchsia), and 3 MIU/8 h (blue)), assuming that CMS renal clearance equals 0.85 CrCL, with CMS non-renal clearance and colistin clearance equal to 50 mL/min.

Effect of altered CMS non-renal clearance on Css,avg

Non-renal elimination of CMS is poorly understood, and it is therefore difficult to speculate on between-patient CMS CLNR variability. However, because CLNR appears in the numerator and denominator of Eqn 4, the effect of altered non-renal CMS elimination on Css,avg should be relatively limited. This was confirmed by simulations conducted for various CLNR values from 50 to 25 and 12.5 mL/min. Furthermore, it is interesting to note that, when CMS non-renal clearance decreases, Css,avg also decreases. There is therefore no risk of toxicity caused by overexposure in case of CMS non-renal clearance impairment (Fig. 5).

Figure 5.

 Simulated colistin concentrations at steady state (Css,avg) as a function of creatinine clearance (CrCL) following a maintenance dose of colistin methanesulphonate (CMS) equal to 3 MIU/8 h and for distinct values of CMS non-renal clearance (50 mL/min (top), 25 mL/min (middle), and 12.5 mL/min (bottom)), assuming that CMS renal clearance equals 0.85 CrCL, and colistin clearance equals 50 mL/min.

Effect of altered colistin clearance on Css,avg

Average colistin concentration at steady state is determined by the previously described pharmacokinetic parameters affecting its rate of formation, and also by its own clearance, as indicated by Eqn 4. Because colistin is presumably characterized by a low hepatic extraction ratio, impaired enzymatic activity is among the factors that may have an effect on its clearance, but experimental data are lacking and probably difficult to obtain. Therefore, simulations are of great interest for predicting the potential effect of reduced colistin clearance on Css,avg. These were conducted for CLcoli values corresponding to 50% and 25% of the estimated value in healthy volunteers, i.e. 25 mL/min and 12.5 mL/min, respectively, and the results are shown in Fig. 6a,b. It is clear from these simulations that impaired colistin elimination may be responsible for relatively high Css,avg values. As an example, Css,avg in the order of 10 mg/mL, which could not be reached in healthy volunteers (Fig. 4), would be obtained in a patient with relatively moderately impaired renal elimination (CrCL = 60 mL/min) but with a relatively more important reduction in CLcoli (12.5 mL/min, corresponding to 25% of the value estimated in healthy volunteers). Therefore, although colistin clearance is presently not well characterized, except for being almost exclusively non-renal, this parameter is critical, not only because high and potentially toxic Css,avg values could be obtained in patients with seriously impaired colistin clearance, but also because it is difficult to predict. It was therefore interesting to compare these simulation results with real-life observations.

Figure 6.

 Simulated colistin concentrations at steady state (Css,avg) as a function of creatinine clearance (CrCL) following various maintenance doses of CMS (1 MIU/8 h (green), 2 MIU/8 h (fuchsia), and 3 MIU/8 h (blue)), assuming that CMS renal clearance equals 0.85 CrCL and CMS non-renal clearance equals 50 mL/min, for colistin clearance equal to 25 mL/min (a) and 10 mL/min (b).

Case reports of unexpectedly high Css,avg in critical-care patients

A first patient treated with multidrug therapy including CMS (3 MIU/8 h) for an infection caused by multidrug-resistant Gram-negative bacteria exhibited generalized seizures, leading a posteriori to colistin plasma concentration determination [23]. CrCL was estimated to be 50 mL/min. According to Fig. 3, i.e. considering that colistin clearance in this patient should be comparable to that in healthy volunteers, Css,avg should be close to 3 mg/mL, when in fact it was determined by LC-MS/MS to be 8.1 mg/mL. This difference could be explained by lower than expected colistin clearance in this patient. Notably, a reduction by half of colistin clearance as compared with that of healthy volunteers (CLcoli reduced from 50 to 25 mL/min) would predict an average steady-state colistin concentration close to the measured value in this patient (Fig. 6a).

A second patient with cystic fibrosis who had received a lung transplant had a much lower CrCL, estimated to be 14 mL/min, indicating considerably reduced renal function. The CMS dosing regimen was fixed at 1.5 MIU/8 h. Once again, according to Fig. 4, Css,avg should be within 2–4 mg/mL. In fact, it was determined to be 10 mg/mL, again suggesting a difficult-to-predict reduction in colistin clearance in this patient.


These recent pharmacokinetic data collected by various groups, which are essentially consistent with each other, raise a number of issues. According to the limited fluctuations in colistin plasma concentration vs. time profiles at steady state, the usual CMS dosing regimen (three times daily) does not seem to be rational. Reducing the number of daily administrations would decrease the workload in critical-care units, reducing the risk of errors, and the difference in terms of colistin concentration vs. time profiles between 3 MIU/8 h and 4.5 MIU/12 h, and to a better degree 9 MIU/24 h, should be relatively limited (Fig. 7b). Therefore, with the same daily dose, two daily administrations could be more appropriate than three. However, CMS concentration vs. time profiles should also be considered. The CMS peak, in particular, would not be very different between the 3 MIU/8 h and 4.5 MIU/12 h schedules, but would be much higher after once-daily administration at the same daily dose, i.e. 9 MIU/24 h (Fig. 7a). Notably, the CMS volume of distribution should determine peak values and should therefore be taken into consideration in this respect. A high CMS peak may have consequences in terms of tolerability and even toxicity. Therefore, although reducing the number of CMS doses appears to be a real possibility on a pharmacokinetic basis, clinical evidence should first be provided before any decision is made.

Figure 7.

 Simulations of steady-state colistin methanesulphonate (CMS) (a) and colistin (b) plasma concentrations vs. time following administration of the same daily dose of CMS with various dosing regimens (9 MIU/24 h (red), 4.5 MIU/12 h (blue), and 3 MIU/8 h (green)), for a typical patient with creatinine clearance set at 60 mL/min, CMS non-renal clearance and colistin clearance equal to 50 mL/min, and CMS and colistin Vss set at 15 L.

Another interesting question concerns the potential benefit of colistin concentration monitoring. The recent pharmacokinetic studies reviewed in this article suggest that colistin concentrations are probably most frequently lower than targeted. This was clear in our ongoing pharmacokinetic study in 33 critical-care patients, in which Css,avg ranged from below the limit of quantification to 6.4 mg/mL, with a median of 1.1 mg/mL. In this study, dosing regimens were selected by clinicians on the basis of their own experience. It appeared that CMS dosing regimens were reduced in patients with mild to moderate renal impairment, when in fact this may not have been appropriate. Therefore, the general message would be to encourage clinicians to change their habits and prescribe higher doses of CMS. However, colistin concentrations in the order of 10 mg/mL may occasionally be observed. Impaired enzymatic activity caused by liver disease or drug–drug interactions could perhaps be responsible for reduced colistin clearance and, in turn, high average steady-state concentrations. Unfortunately, these phenomena are mostly unknown, and colistin clearance impairment in a particular individual is almost impossible to predict at present. Therefore, colistin concentration monitoring may be advisable to provide security for clinicians when, at the same time, they are encouraged to increase CMS dosing regimens.

Because colistin concentration fluctuations are limited within a dosing interval, especially when CMS is given three times daily, we believe that one should not try to differentiate between peak and trough concentrations. However, although a plasma concentration determined at any time would probably provide an appropriate estimate of Css,avg for the purposes of drug monitoring, it would be more advisable to sample immediately before CMS dosing, not only because this is frequently more convenient from a practical view, but also, and more importantly, because CMS concentrations would then be minimal and the risk of colistin concentration overestimation resulting from post-sampling CMS hydrolysis would therefore be considerably reduced. Colistin drug monitoring was recently introduced in our institution, using a two-step procedure with colistin concentration determination after 2 or 3 days of treatment to estimate CLcoli in the patient and readjust the CMS dosing regimen. Most often, clinicians expected colistin concentrations that were higher than observed. This is particularly true for clinicians who are used to relying on microbiological assays for colistin, because, with the use of these non-specific assays, artificially high and misleading ‘peak concentrations’ are obtained.

In conclusion, tremendous progress has been made recently in our understanding of the complex colistin pharmacokinetics following intravenous dosing with CMS in critical-care patients. It is not yet possible to predict colistin concentrations in every individual patient with total accuracy. However, complementary studies are being conducted by various groups, and rapid progress is expected. The next step will be to better understand colistin pharmacokinetics–pharmacodynamics, and to characterize drug–drug interactions both at a pharmacokinetic and at a pharmacokinetic–pharmacodynamic level, as critical-care patients infected by multidrug-resistant bacteria are not treated with colistin alone, but with a mixture of several antibiotics. Therefore, the challenge is real, but the fog is lifting.


This study was supported by internal funding.


The authors thank Dr. H. Derendorf for his suggestion concerning the title of this article.

Transparency Declaration

Nothing to declare.