Colistin monotherapy vs. combination therapy: evidence from microbiological, animal and clinical studies
Corresponding author and reprint requests: M. E. Falagas, Alfa Institute of Biomedical Sciences (AIBS), 9 Neapoleos Street, Marousi 151 23, Greece
Colistin is commonly the last resort for treatment of infections caused by multidrug-resistant Gram-negative bacteria. In clinical practice, it is frequently used as combination therapy in order to improve its antibacterial activity, despite the consequent increase in toxicity. The available evidence from various studies (microbiological, animal and clinical studies, retrieved from the PubMed and Scopus databases) regarding the comparative effectiveness of colistin monotherapy and colistin combination therapy was evaluated. Most of the microbiological studies examined colistin monotherapy vs. combinations with rifampicin (nine studies) or carbapenems (three studies) for Pseudomonas aeruginosa or Acinetobacter baumannii infections. A synergistic effect was detected in all the studies examining the combination of colistin and rifampicin, whereas carbapenems exhibited a synergistic effect in two of three studies. Most of the animal studies examined colistin monotherapy vs. combinations with rifampicin, carbenicillin, piperacillin and imipenem for treatment of P. aeruginosa, A. baumannii or Escherichia coli infections. Mortality rates were significantly lower in the combination treatment arm in three of six relevant studies. However, data from the small number (four) of relevant human studies suggest non-inferiority of colistin monotherapy as compared with combination therapy. In conclusion, microbiological studies suggest superiority of colistin combination treatment, which is in contrast to preliminary data from studies in humans. Results from animal study data are equivocal. There is an urgent need for appropriately designed and powered clinical trials addressing this apparently controversial situation.
The mounting prevalence worldwide of infections caused by multidrug-resistant (MDR) Gram-negative bacteria, in particular Pseudomonas aeruginosa, Acinetobacter baumannii and Klebsiella pneumoniae, is causing substantial concern . The lack of novel antimicrobials for Gram-negative infections under development, especially for the treatment of infections due to P. aeruginosa, has forced clinicians to reappraise the clinical value of colistin, a polymyxin antibiotic discovered c. 60 years ago [2–5].
Colistin is a multicomponent polypeptide antibiotic, composed mainly of colistin A and colistin B; its use was limited by its renal toxicity, and it was replaced in the 1970s by antibiotics considered to be less toxic. Pharmacokinetic and pharmacodynamic information on colistin is limited, perhaps due to the moderate clinical interest in polymyxins during the 1980s and 1990s and the difficulties in accurately measuring colistin and colistimethate sodium separately in biological samples .
More recently, colistin has increasingly been used as salvage therapy [2,6] in combination with one or more antibacterials for the treatment of severe infections in critically ill patients . Despite the fact that polymyxins have been available for over 50 years, in vitro pharmacodynamic, animal or clinical studies regarding the pharmacokinetic parameters (maximal concentration (Cmax)/MIC, area under the curve/MIC (AUC/MIC), and proportion of time/MIC (%T/MIC)) of colistin formulations are scarce . In clinical practice, colistin is usually administered at 8-h intervals.
The medical community has shown great interest in the comparative benefit, if any, of combination therapy vs. monotherapy in the treatment of Gram-negative and Gram-positive bacterial infections. Several studies have questioned the clinical superiority of combination therapy and have addressed the issues of the increased toxicity and the increased cost related to combination treatment regimens [9–13]. However, in severe infections caused by MDR Gram-negative pathogens, most clinicians would currently have reservations about colistin monotherapy. Moreover, a major concern regarding colistin monotherapy is the potential problem of heteroresistance among Gram-negative bacterial populations exposed to colistin alone [8,14]. Currently, there is a dearth of data regarding the clinical value of combination therapy and the clinical impact of heteroresistance. An overview of data from in vitro, animal and clinical studies regarding the efficacy and effectiveness of colistin combinations vs. colistin alone is presented in this review. This brief synopsis of relevant data may contribute to an evaluation of whether the superiority of colistin combinations, as compared with colistin alone, as depicted in in vitro studies, is maintained and verified in animal and, most importantly, in clinical studies.
Microbiological studies comparing colistin monotherapy and combination therapy
In vitro studies comparing colistin alone and colistin in combination with other antibiotics were considered to be relevant to the focus of this review. A search of the PubMed and Scopus databases yielded 16 relevant studies [15–30]. The search terms used were colistin, polymyxin or colimycin, in vitro study, Acinetobacter, Pseudomonas, Stenotrophomonas maltophilia, and Klebsiella. One study  was excluded because it referred to polymyxin B. Data retrieved from relevant studies are presented in Table 1.
Table 1. Summary of microbiological studies comparing colistin monotherapy and combination therapy
|Tascini et al. ||1998/Italy||MIC determination, checkerboard assay||Acinetobacter baumannii (5)||COLa alone, COL + RIF, COL + AMP/SUL||All strains were susceptible to COL (MIC: 0.5–1 mg/L), and two strains had intermediate susceptibility to RIF and three were resistant. COL + RIF was synergistic against three of five strains. COL + AMP/SUL was indifferent in all five strains|| |
|Hogg et al. ||1998/UK||MIC determination, checkerboard assay||A. baumannii (13)||COLa alone, COL + RIF||All strains were susceptible to COL (MIC: 0.03–4 mg/L), and 11 were resistant to RIF. COL + RIF was synergistic against 11 strains (FIC indexd: 0.07–0.63) and indifferent against two (FIC index: 1 and 1.13, respectively); antagonism was not observed|| |
|Rynn et al. ||1999/UK||MIC determination, time–kill assay||Pseudomonas aeruginosa||COLb alone (at concentrations of 0.5 and 5 mg/L), COL + CTX/AZI/MER/GEN/PIP/CIP||For COL (MIC: 2 mg/L) + CTX, COL + AZT, COL + MER and COL + CIP, the pattern for all of the combinations (high or low concentrations) was to produce smaller AUBKCs than single agents||These studies indicate that addition of COL to other antipseudomonal drugs tends to produce smaller AUBKCs and hence greater killing of P. aeruginosa than monotherapy|
|Giamarellos-Bourboulis et al. ||2001/Greece||MIC determination, time–kill assay||A. baumannii (39)||COLc alone, RIF alone, COL + RIF||All strains were resistant to AMP/SUL, CTX, CEF, CTD, CFP, AMI, and CIP, and susceptible to COL (MIC: 0.03–2 mg/L). IMI inhibited 89.7% of strains, and RIF 15.4%. Synergy was found between 1× MIC of COL and RIF in six strains (15.4%) at 6 h of growth and in 20 strains (51.3%) at 24 h of growth. The respective results for synergy between 4 × MIC of COL and RIF were six strains (15.4%) and 26 strains (66.7%). No synergy was found at 2 and 4 h of growth||The in vitro activity of COL was greatly increased in the presence of RIF. Synergy between COL and RIF was dependent on the applied concentration of COL. COL + RIF prevented regrowth at 6 h, as occurred with COL alone|
|Giamarellos-Bourboulis et al. ||2002/Greece||MIC determination, time–kill assay||CMZ/RIF-resistant Stenotrophomonas maltophilia (24)||COLc alone, RIF alone, CMZ alone, COL + RIF, COL + CMZ||MIC50 and MIC90 of COL were 4 and 16 mg/L respectively. Synergy of COL and RIF involved approximately 60% of strains, and it was expressed over the first hours of exposure. Synergy of colistin and trimethoprim–sulphamethoxazole was found when colistin was applied at a concentration of 4 × MIC, involving 25% of isolates after 6 h of growth and 41.7% of strains after 24 h of growth||COL + RIF and COL + CMZ prevented regrowth occurring with colistin monotherapy after 24 h of growth|
|Giamarellos-Bouboulis et al. ||2003/Greece||MIC determination, time–kill assay||MDR P. aeruginosa (17).||COLc + RIF||The MIC of COL was <8 mg/L for eight strains, and ≥8 mg/L for nine strains. Synergy between COL and RIF was found in four, six, seven, and two strains after 2, 4, 6 and 24 h of growth respectively.||Synergy between COL and RIF was documented in four (50%) of eight isolates with a COL MIC <8 mg/L, and in three (33.3%) of nine with COL MIC ≥8 mg/L.|
|Gunderson et al. ||2003/USA||MIC determination, time–kill assay||MDR P. aeruginosa (2)||COLc alone, COL + CTD, COL + CIP||The MIC of COL for each strain was 0.125 mg/L. COL + CTD was synergistic at 24 h. Adding CIP to COL did not enhance antibiotic activity||The combination of COL with a Cmax of 18 mg/L and CTD was only slightly superior to the combination with COL with a Cmax of 6 mg/L, thus suggesting that the antibacterial effect of COL combined with CTD can be maximized at a peak concentration of <18 mg/L|
|Tascini et al. ||2004/Italy||MIC determination, checkerboard and time–kill assays||MDR P. aeruginosa strains (7)||COLb alone, RIF alone, COL + RIF||COL alone had no bactericidal effect, and after 6 h was unable to counteract bacterial growth; with the addition of RIF, COL became bactericidal and the effect was prolonged for 12 h. RIF + COL was fully (one strain) or partially (five strains) synergistic in six of seven strains, and MICs in the combination arm were reduced to easily obtainable therapeutic levels||The time–kill curves showed that the combination was bactericidal against the strains tested|
|Hill et al. ||2005/Canada||MIC and MCBT determinations||MDR P. aeruginosa (16)||COLc alone, COL + CIP/CMZ/CTD/AZI||COL (32 mg/L) was bactericidal most often, killing all 16 (100%), 12 of 16 (75%) and 4 of 16 (25%) of strains tested under aerobic, anaerobic and biofilm conditions, respectively. COL combinations enhanced bactericidal activity under anaerobic and biofilm conditions||As COL was effective against all isolates as a single agent, it was not surprising that COL combinations were the most effective under all conditions|
|Giacometti et al. ||2005/Italy||MIC determination, checkerboard assay||MDR P. aeruginosa (20), MDR S. maltophilia (20)||COLa alone, COL + histatin derivative P-113||All strains of S. maltophilia were susceptible to COL (MIC: 0.5–4 mg/L). The MIC for P. aeruginosa was 0.5–8 mg/L. For the combination, the FIC indexe was 0.927 and 0.917 for P. aeruginosa and S. maltophilia, respectively||The combination of the histatin derivative P-113 and COL showed an indifferent effect|
|Timurkayanak et al. ||2006/Turkey||MIC determination, checkerboard assay||MDR A. baumannii (5), MDR P. aeruginosa (5)||COLc alone, AZI alone, DOX alone, RIF alone, MER alone, COL + AZI/DOX/RIF/MER||All strains were susceptible to COL alone. COL + RIF was fully synergistic against four of the A. baumannii strains tested. COL + MER and COL + AZI were synergistic against three of these five strains each. For the P. aeruginosa strains, the only regimen that was synergistic was COL + RIF (two strains only)||COL + DOX was partially synergistic against four of the A. baumanii strains. COL + MER, COL + AZI and COL + DOX showed either partial synergy, or additive or indifferent effects, on MDR P. aeruginosa strains|
|Cirioni et al. ||2007/Italy||MIC determination, checkerboard and time–kill assays||P. aeruginosa ATTC 27853, MDR P. aeruginosa||Tachyplesin III, COLc, IMI, tachyplesin III + IMI, COL + IMI||COL exhibited MICs of 4 mg/L for P. aeruginosa ATCC 27853 and 8 mg/L for the multiresistant strain, respectively. Synergy was observed between COL and IMI for both P. aeruginosa ATCC 27853 (FIC indexe: 0.385) and the clinical isolate (FIC indexe: 0.458)||These data were confirmed by the time–kill synergy studies|
|Song et al. ||2007/Korea||MIC determination, time–kill assay||MDR A. baumannii (43)||COLc, COL + RIF||All strains were susceptible to COL (MIC range 0.5–4 mg/L). The COL time–kill assay demonstrated bactericidal activity against A. baumannii at concentrations of 4 × MIC and 8 × MIC. COL + RIF was synergistic and bactericidal at 1 × MIC||Synergistic and bactericidal effects of COL + RIF were sustained at the MIC level for 24 h|
|Li et al. ||2007/Australia||MIC determination||A. baumanii strains (8)||COLa + RIF||Strains were susceptible to COL (MIC range 0.25–2 mg/L). The FIC indexe ranged from 0.14 to 0.53||COL + RIF was synergistic. COL + RIF was tested against COL-resistant strains as well, but there was no growth, so the FIC index was not calculated|
|Tan et al. ||2007/Singapore||MIC determination, time–kill assay, Etest study||IMI-resistant A. baumannii (13)||COLa, MIN, COL + MIN||All strains were susceptible to COL (MIC range 0.5–2 mg/L), whereas 70% were susceptible to MIN. At 1 × MIC, neither COL nor MIN when tested alone demonstrated bactericidal activity. COL + MIN was rapidly bactericidal. Synergy was detected in 92% of isolates at 24 h (also in75% of MIN-resistant strains). Etest showed no synergy (FIC indexe ranged between 1.1 and 1.5)||Bacterial regrowth at 24 h for both antibiotics, when tested singly. For COL + MIN, there was minimal evidence of bacterial regrowth at 24 h, when a ≥3 log10 reduction in CFU was obtained. No agreement between time–kill and Etest methods for synergy testing|
In all these studies, susceptibility to colistin was previously evaluated by determination of MICs. The activity of the combination of colistin and other agents was evaluated using the checkerboard microbroth dilution method in six studies [15,16,23,25–27]. Concentration–time–kill curves were used in nine studies [17–19, 21–23,27,28,30]. In one investigation, an Etest study was also performed .
The antimicrobial agents combined with colistin were rifampicin [15,16,18–19,21,23,26,28,29], azithromicin [17,24,26], imipenem , meropenem [17,26], gentamicin , piperacillin , ciprofloxacin [17,22,24], co-trimoxazole [19,24], ceftazidime [22,24], doxycycline , minocycline , and the histatin derivative P-113 , i.e. a combination of polypeptides. However, the antimicrobials most frequently combined with colistin were rifampicin (ten studies) and carbapenems (four studies).
The most commonly studied organism was P. aeruginosa (8/13 studies [17,21–27]). In one study, Rynn et al. found that the addition of colistin to other antipseudomonal agents produced a smaller area under the bactericidal killing curves, and hence a greater killing effect on P. aeruginosa, than monotherapy . This effect was independent of the colistin concentration (0.5 and 5 mg/L, respectively). In another study, two MDR strains, which were colistin-susceptible, were studied using a time–kill assay . The combination of colistin and ceftazidime was synergistic, with only slight superiority of the combination with colistin at Cmax 18 mg/L as compared to the combination with colistin at Cmax 6 mg/L. Ciprofloxacin added to colistin did not enhance antibiotic activity.
Tascini et al. compared colistin plus rifampicin with colistin alone in seven MDR P. aeruginosa strains on which colistin alone had no bactericidal effect and after 6 h was unable to counteract bacterial growth . They found that, in combination with rifampicin, colistin became bactericidal and the effect was prolonged for 12 h. Moreover, the combination resulted in synergy in six of seven strains. Synergy between colistin and rifampicin was also confirmed by Timurkaynak et al. in five MDR A. baumannii strains .
However, results were equivocal when carbapenems were added to colistin [26,27] in the case of MDR P. aeruginosa. Indeed, in one study , colistin plus meropenem was additive in two of five strains and indifferent in the remaining three strains. In the other study of one MDR strain with an MIC of colistin of 8 mg/L, synergy was observed between colistin and imipenem with a fractional inhibitory concentration index of 0.458 .
Regarding the activity under anaerobic and biofilm conditions, Hill et al.  found that colistin combinations with ciprofloxacin, co-trimoxazole, ceftazidime or azithromicin enhanced bactericidal activity against 16 MDR P. aeruginosa strains (all susceptible to colistin alone). Finally, Giacometti et al.  demonstrated that colistin plus a combination of peptides (histatin derivative P-113) had an indifferent effect.
Seven in vitro studies of A. baumannii were found [15,16,18,26,28–30]; in all of them, colistin was compared to colistin plus rifampicin, and in one study , other antimicrobials (azithromicin, doxicycline, and meropenem) were also studied in combination with colistin. In the study conducted by Hogg et al., all strains were susceptible to colistin, and 85% (11/13) were resistant to rifampicin . The combination of colistin plus rifampicin was synergistic against 85% of isolates and indifferent in the remaining 15%, whereas no antagonism was found. In the study conducted by Giamarellos-Bourboulis et al. , which used time–kill curve assays, 39 MDR A. baumannii strains (all susceptible to colistin) were studied. The authors found that the in vitro activity of colistin was highly increased in the presence of rifampicin, and the synergy was dependent on the concentration of colistin (synergy at 24 h of growth was 51.3% and 66.7% at 1 × MIC and 4 × MIC of colistin, respectively). In the study conducted by Timurkaynak et al. , five MDR A. baumannii strains (all susceptible to colistin) were studied, using the checkerboard methodology. The combination with rifampicin was fully synergistic against four of five strains, whereas the combinations with meropenem and azithromicin were synergistic against three of five strains each. Interestingly, the combination with doxycycline was partially synergistic against four of five strains.
A recent study conducted by Song et al.  showed that colistin alone exhibited a bactericidal effect only at high concentrations (above 4 × MIC), which is clinically important, considering the dose-dependent nephrotoxicity of colistin. However, they found that colistin plus rifampicin had synergistic and bactericidal effects on the MIC level that were sustained for more than 24 h.
Similarly, Tan et al.  found that at 1 × MIC, neither colistin nor minocycline, when tested alone, demonstrated considerable bactericidal activity against MDR-resistant A. baumannii strains, and that there was also evidence of bacterial regrowth at 24 h in the presence of both antibiotics, when tested alone. The use of colistin and minocycline in combination showed a rapidly bactericidal and synergistic effect, with minimal evidence of bacterial regrowth at 24 h, when a ≥3 log10 reduction in CFU was obtained. Interestingly, the Etest results were in disagreement with those of the time–kill assay.
Two studies comparing the activity of colistin combination therapy with that of other antimicrobial agents against S. maltophilia were found [19,25]. In one of them, 24 co-trimoxazole/rifampicin-resistant strains with MIC50 and MIC90 for colistin of 4 and 16 mg/L, respectively, were tested using colistin alone or in combination with rifampicin or co-trimoxazole . Synergy between colistin and rifampicin was observed in 60% of the isolates, and occurred in the first hours of exposure. Synergy with co-trimoxazole was found at a colistin concentration of 4 × MIC (25% of the isolates at 6 h and 41.7% at 24 h of growth). In conclusion, both interactions prevented regrowth occurring under colistin monotherapy after 24 h of growth. Finally, in another study, Giacometti et al.  found an indifferent effect of the histatin derivative P-113 combined with colistin on 24 MDR S. maltophilia strains.
Animal studies comparing colistin monotherapy and combination therapy
Animal studies comparing the effectiveness of colistin combination therapy with monotherapy were considered to be relevant to the focus of this review. The search of the PubMed and Scopus databases yielded six relevant studies [27,31–35]. The search terms used were colistin or colimycin, in vivo study, animal study, Acinetobacter, Pseudomonas, S. maltophilia, and Klebsiella. Data retrieved from the relevant studies are presented in Table 2.
Table 2. Summary of animal studies comparing colistin monotherapy and combination therapy
|Saslaw et al. ||1973/USA||Experimental rhesus monkey models of sepsis||Pseudomonas aeruginosa||Controls (10), COLa (10), carbenicillin (10), COL + carbenicillin (12)||Mortality at 5 days: COL (6/10), carbenicilin (3/10), controls (8/10)||COL + carbenicillin (5/12)||The response of monkeys treated with the antibiotic combination did not differ significantly from that of monkeys treated with a single agent|
|Giacometti et al. ||2003/Italy||Experimental intraperitoneal infections in adult Wistar rats. Controls: uninfected Wistar rats||Escherichia coli ATCC 25922||Placebo (15), PIP (15), TEI (15), VAN (15), COLb (15), buforin II (15), TEI + PIP (15), VAN + PIP (15), COL + PIP (15), buforin II + PIP (15)||Lethality (48 h): uninfected/no treatment (0/15), infected/no treatment (14/15), PIP (5/15), TEI (14/15), VAN (15/15), COL (5/15), buforin II (4/15), TEI + PIP (3/15), VAN + PIP (5/15), buforin II + PIP (0/15)||Lethality (48 h): 0/15||Only COL and buforin II combined with PIP significantly decreased mortality as compared to COL or PIP alone. COL and buforin II combined with PIP produced the greatest suppression of bacterial growth|
|Montero et al. ||2004/Spain||Mouse pneumonia model. Lung bacterial counts (log10 CFU/g) after 48 h of therapy||Two strains of carbapenem-resistant Acinetobacter baumannii||Controls (33), IMI, sulbactam, TOB, RIF, COLb were used in monotherapy and in combination therapy (four mice per group). COL was combined only with RIF||Lung bacterial count for high carbapenem-resistant A. baumannii: controls (10.82 ± 0.33), RIF (5.62 ± 0.26), COL (8.38 ± 1.22), IMI (11.01 ± 0.2), RIF + IMI (3.79 ± 0.99), RIF + TOB (3.96 ± 0.30)||RIF + COL (5.59 ± 1.17)||RIF + COL had the same effectiveness as RIF alone regarding lung bacterial counts. For strains of A. baumanii that are highly resistant to IMI, a combination of RIF with IMI, TOB or COL may be useful, if resistance to RIF is moderate|
|Pantopoulou et al. ||2007/Greece||Thigh infection model in 86 neutropenic Wistar rats. Survival was recorded in ten animals of each group||MDR A. baumannii||Controls (20), RIF (20), COLa (22), RIF + COL (20)||Median survival: controls (2 days), RIF (2.5 days), COL (4 days). Mortality rates after 6 days: controls (10/10), RIF (10/10), COL (10/10)||Median survival: 4 days. Mortality rate after 6 days: 7/10||Survival with COL (alone or in combination with RIF) was significantly higher vs. controls. Mortality rates on the sixth day of follow-up showed the enhancement of the effect of COL by the addition of RIF. Statistically significant decreases in numbers of bacteria were found in blood and liver of the combined group vs. controls|
|Cirioni et al. ||2007/Italy||Mouse model of sepsis. Main outcome: mortality, quantitative blood culture and detection of lipopolysaccharide, TNF-α and IL-6 plasma levels||MDR P. aeruginosa||Controls (20), tachyplesin (20), COLb (20), IMI (20), tachyplesin and IMI (20), IMI + COL (20)||Overall deaths: tachyplesin (6/20), COL (6/20), IMI (16/20), controls (20/20)||Overall deaths: COL + IMI (3/20), tachyplesin + IMI (2/20)||Combination treatment groups had significantly lower mortality and bacteraemia than monotherapy treatment groups. Tachyplesin III and COL treatments (alone or in combination) resulted in marked decreases (p <0.05) of endotoxin, TNF-α, and IL-6 plasma levels as compared to those of controls|
|Cirioni et al. ||2007/Italy||Experimental Wistar rat models of sepsis. Prospective, randomized, controlled animal study||P. aeruginosa||Controls (15), RIF (15), COLb + RIF (15)||Mortality at 72 h (MDR strain): controls (15/15), RIF (15/15), COL (9/15)||Mortality at 72 h: COL + RIF (4/15)||Combination of COL and RIF resulted in a significant reduction in bacterial count as compared with COL monotherapy, even though no significant difference was found in positive blood cultures and mortality rates between the two groups|
P. aeruginosa, A. baumanii and Escherichia coli were tested in three studies [27,31,35], two studies [33,34], and one study , respectively. Experimental mouse models were adopted in three studies [27,31,33] and rat models in the remaining three [32,34,35]. Colistin was combined with rifampicin in three studies [33–35], and with imipenem , carbenicillin  or piperacillin  in the remaining three studies.
With regard to experimental P. aeruginosa infections, in an old study, Saslaw et al.  found that the mortality rate in septic monkeys treated with colistin alone (6/10) did not differ significantly from that in septic monkeys treated with colistin plus carbenicillin (5/12). In two more recent Italian studies [27,35] using an experimental model of sepsis in rats, Cirioni et al.  found that: (1) overall mortality in animals treated with colistin alone was 30% (6/20) vs. 15% (3/20) with colistin plus imipenem; and (2) mortality at 72 h was 60% (9/15) in animals treated with colistin alone and 27% (4/15) in animals treated with colistin plus rifampicin . The latter study was a prospective, randomized, controlled animal study, and the authors found that treatment with colistin 1 mg/kg plus rifampicin 10 mg/kg resulted in a significant reduction in bacterial count in peritoneal fluid as compared with colistin alone, even though no significant differences were found in mortality rates between the two groups.
In both of the two experimental MDR A. baumannii model studies, colistin alone was compared with colistin plus rifampicin. In the first study, which used a mouse model of pneumonia, the endpoint was the lung bacterial count expressed as log10 CFU/g, mean ± SD, with 8.38 ± 1.22 for colistin alone and 5.59 ± 1.17 for colistin plus rifampicin; the differences were not statistically significant . In the second study, which used a thigh infection model in neutropenic Wistar rats, the median survival rate was 4 days and the mortality rate after 6 days was 100% (10/10) with colistin alone, whereas the median survival was 4 days (the difference between colistin and colistin plus rifampicin was not statistically significant) and the mortality rate after 6 days was 70% (7/10) with colistin plus rifampicin (p 0.018 between groups) .
Finally, in the only study of E. coli (strain ATCC 25922), Giacometti et al.  used an experimental intraperitoneal infection model in adult Wistar rats, with uninfected rats as controls. Mortality with colistin alone at 48 h was 33.3% (5/15) vs. 0% (0/15) in rats treated with colistin (1 mg/kg) plus piperacillin (60 mg/kg). Similarly, the intraperitoneal bacterial cell count was 3.4 × 103 CFU/mL ± (0.9 × 103) with colistin alone vs. <10 CFU/mL with the combination regimen, whereas the endotoxin level was almost the same in the two conditions.
Clinical studies comparing colistin monotherapy and combination therapy
Clinical studies comparing the effectiveness of colistin combination therapy with monotherapy were considered to be relevant to the focus of this review. Only four relevant studies were identified by searching the PubMed and Scopus databases. The search terms used were colistin or colimycin in combination with monotherapy, intravenous, Acinetobacter, Pseudomonas and Klebsiella [36–39]. Data retrieved from relevant studies are presented in Table 3.
Table 3. Summary of clinical studies comparing colistin monotherapy and combination therapy
|Tascini et al. ||2006/Italy||Retrospective study||Diabetic foot infection with/without osteomyelitis||MDR Pseudomonas aeruginosa||Colistin alone (1 MIU every 12 h; four patients) vs. colistin (1 MIU every 12 h) + rifampicin (three patients) or imipenem (one patient)||3/4||2/4||Renal insufficiency: one patient in the monotherapy group||No difference in response and safety rates|
|Falagas et al. ||2006/Greece||Retrospective study||Pneumonia, urinary tract infection, intra-abdominal infection, spondylodiskitis, surgical/skin and soft tissue infection, bacteraemia, catheter-related infection||Acinetobacter baumanii, P. aeruginosa, Stenotrophomonas maltophilia, Enterobacter cloacae, Escherichia coli||Colistin alone (4.6 ± 2.3 MIU/day; 14 patients) vs. colistin (4.6 ± 2.3 MIU/day) + meropenem (57 patients)||12/14||39/57||Nephrotoxicitya: four patients in the combination group||No difference in response and nephrotoxicity rates. Survival was significantly higher in patients treated with colistin monotherapy (p 0.007)|
|Linden et al. ||2003/USA||Prospective study||Pneumonia, bacteraemia, wound, intra-abdominal infection, endocarditis||MDR P. aeruginosa||Colistin alone (dose ranging from 1 to 5 mg/kg/day according to renal function; ten patients) vs. colistin (dose ranging from 1 to 5 mg/kg/day according to renal function) + amikacin (four patients) or antipseudomonal β-lactam (nine patients)||6/10||8/13||ND||No difference in response rates|
|Conway et al. ||1997/UK||Prospective study||Respiratory exacerbations in patients with cystic fibrosis||P. aeruginosa||Colistin alone (2 MIU every 8 h; 36 patients) vs. colistin + aztreonam, azlocillin, piperacillin, ceftazidime, imipenem, or ciprofloxacin (35 patients)||36/36||35/35||Mild neurological symptoms: 33/36 in the monotherapy group, 36/36 in the combination group. Severe neurological symptoms in 1/36 patients in the monotherapy group. Statistically significant decrease in creatinine clearance in the combination therapy group as compared with baseline. Significant increase in urea in both groups as compared with baseline values||Statistically significant increase in patients with normal C-reactive protein in the combination therapy group. Differences in respiratory function tests and clinical scores between groups were not statistically significant by day 12 of treatment|
The overall number of patients examined in the monotherapy treatment arm was 46, whereas the number of patients in the comparator group was 48. The antimicrobial agents used in combination with colistin were rifampicin, imipenem, meropenem, aztreonam, azlocillin, piperacillin, ceftazidime and ciprofloxacin. The effectiveness of the various regimens was examined in diabetic foot infection , pneumonia ([37,38]), bacteraemia [37,38], intra-abdominal infection ([37,38]), endocarditis , urinary tract infection , spondylodiskitis , soft tissue infection , and respiratory exacerbations in patients with cystic fibrosis . The responsible pathogens were MDR P. aeruginosa [35–39] and A. baumannii .
In all four studies identified there were no statistically significant differences in the effectiveness and safety rates. However, a retrospective study comparing colistin monotherapy and combination therapy with meropenem found a higher survival rate in the monotherapy treatment arm after adjustment for the variables for which significant differences were noticed . Another study that examined colistin effectiveness in patients with cystic fibrosis demonstrated no statistically significant differences between treatment groups except for higher C-reactive protein normalization rates in the combination therapy group by day 12 of treatment .
Non-comparative studies of colistin monotherapy and combination therapy
As the emergence of MDR strains susceptible only to colistin poses an important clinical problem, an increasing number of studies have sought to assess the clinical effectiveness of colistin. Most of these studies have been non-comparative and have examined either colistin combination regimens [7,40–43] or colistin monotherapy [44,45]. In addition, a number of comparative studies [46–48] have examined the effectiveness of colistin regimens vs. non-colistin regimens.
A review of the data from noncomparative studies concerning the clinical effectiveness of colistin monotherapy and combination therapy suggests that therapy with colistin alone achieves cure and/or improvement rates ranging from 57% to 78%, whereas the equivalent rates for combination therapy are 67–74%. Despite the very high similarity between colistin monotherapy and combination therapy in rates of effectiveness, the interpretation of existing data is hampered by the considerable variability in settings, dosing regimens and patient populations among studies. Notably, in a prospective study conducted by Reina et al. , the rates of improvement reported with colistin monotherapy, although comparable to those achieved in the non-colistin group, were substantially lower than those reported by the rest of the relevant studies. Factors contributing to this discordance include the facts that criteria for improvement were strictly defined as simultaneous normalization of central temperature (38°C or less), white blood cell count (10 000/mm3 or less) and PaO2/FIO2 (greater than 187) and that the course of infection was evaluated on day 6, which may be too early for the examined parameters to show improvement.
Evaluation of the available evidence
From a review of the available literature, it is obvious that there have been only a limited number of studies comparing the clinical effectiveness of colistin monotherapy and combination therapy in patients with or without cystic fibrosis. Moreover, the number of patients participating in the existing studies is markedly low, and additional limitations stem from the fact that half of the studies are retrospective and there is heterogeneity in the definition of outcome, variability in the dosing regimens and differences in the susceptibility testing methods (disk diffusion or broth dilution).
Despite the evidence of synergy between colistin and various antimicrobial agents (imipenem, rifampicin, ceftazidime) in a proportion of in vitro and animal studies, the evidence from the limited clinical studies available suggests that colistin combination therapy is not superior to monotherapy. Also, data from a small retrospective study  conducted in Greece support an association between colistin monotherapy and better rates of survival that remained statistically significant after adjustment for various relevant variables. Although this study had limitations inherent to the study design of retrospective studies, it is interesting that the favourable result with respect to colistin monotherapy remained statistically significant even after adjustment for important clinical characteristics, with differences being seen in the distribution between the two groups, i.e. site of infection and responsible pathogen. However, it should be emphasized that the favourable outcome of monotherapy reported in this study may be due to the effect of other confounding factors not adjusted for in the multivariate analysis.
It seems that the severity and life-threatening character of infections caused by MDR pathogens has deterred clinicians from providing colistin monotherapy in a considerable proportion of cases. In addition, it should be emphasized that there have not been recent randomized controlled trials examining the various aspects of the effectiveness of colistin, including the comparison of colistin monotherapy with combination therapy.
Although colistin might be the last resort for severe infections in critically ill patients, as it exhibits antimicrobial activity against MDR Gram-negative pathogens, there is substantial evidence that it does not escape resistance. Li et al. , in 2006, described the emergence of A. baumanii subpopulations resistant to colistin, and A. baumanii heteroresistance to carbapenems has been detected by Etest in strains isolated in Greece . As an increase in the daily dosage of colistin in order to eradicate resistant subpopulations could confer higher toxicity rates, the use of combination regimens has been proposed. However, it is noteworthy that combination treatment has been suggested for colistin despite the fact that A. baumanii heteroresistance to carbapenems has also been described. Moreover, the clinical implications of heteroresistance have not been elucidated.
It should be emphasized that the potential effect of antimicrobial heteroresistance on clinical outcomes in patients with infections requires further study . Some experts hold that heteroresistance is an under-appreciated phenomenon in both diagnosis and treatment of infectious diseases. Also, available data concerning heteroresistance have provided some scientific arguments in support of the possible advantages of combination regimens in eradicating resistant subpopulations .
Considering the preliminary data from clinical studies that support the non-inferiority of colistin monotherapy as compared with combination therapy, the increased toxicity that combination regimens may have, and the inadequate evidence from animal studies that there is definitive synergy between colistin and selected antimicrobial agents, we believe that there is an urgent need for clinicians in various parts of the world to share their experience regarding the comparative effectiveness and safety of colistin monotherapy and colistin combination therapy, even in the form of retrospective studies. In addition, and most importantly, we believe that the design and performance of appropriately designed randomized controlled trials comparing the effectiveness and safety of colistin monotherapy with that of colistin combination therapy would clarify the existing uncertainty. Also, further reporting of the development of resistant strains during colistin treatment would help to answer the question of whether the heteroresistance detected in in vitro studies translates into an increased threat of resistance in clinical practice when using colistin monotherapy.
The authors declare no sources of funding and no conflicts of interests.