Colistin is not active against any Gram-positive or anaerobic bacteria. It is primarily used for invasive infections caused by XDR infections, particularly Acinetobacter spp., Pseudomonas spp., Stenotrophomonas spp., Klebsiella spp., and other Enterobacterales. These are typically hospital-acquired infections.

It is also not active against some Gram-negative bacteria for example S. marcescens, Proteus spp., Morganella spp., P. stuartii and B. cepacia complex.

In general, colistin should be initiated based on confirmed or presumptive identification of an invasive XDR Gram-negative infection where no other suitable agent is available. Where possible, treatment with colistin should be in consultation with neonatologist and/or infectious diseases subspecialist. Presumptive identification refers to scenarios where the precise identification and antimicrobial susceptibility profile of an invasive isolate is underway and the patient is in a unit where carbapenem-resistant Gram-negative pathogens are prevalent.

Colistin should not be used for non-invasive or superficial infections or to treat colonisation. Clinical signs and symptoms accompanied by identification of an XDR Gram-negative pathogen from a normally sterile site (e.g. blood, cerebrospinal fluid, synovial fluid, serosal fluid) is sufficient justification for colistin treatment. Conversely, identification of such a pathogen from other, not normally sterile sites (e.g. tracheal aspirate, sputum, skin swab, stool) generally should not prompt colistin treatment unless there is compelling clinical, radiological or laboratory evidence suggesting invasive disease – such cases should be discussed with a clinical microbiologist, neonatologist and/or infectious diseases subspecialist before colistin is initiated.

See the following recommendations regarding empiric use of colistin

Broth microdilution is the only reliable method for determining colistin MICs according to European Committee on Antimicrobial Susceptibility Testing (EUCAST).^12,13 The Clinical and Laboratory Standards Institute (CLSI) includes broth disk elution and agar dilution on their list of approved methods. This requires specialised equipment and expertise, and laboratories unable to perform MICs using this method, should refer isolates to laboratories with capacity to perform such testing. Disk diffusion and gradient diffusion methods should not be performed, as these are unreliable methods for colistin MIC determination. Broth microdilution is the only approved method for polymyxin B.

When the colistin MIC is ≥ 2 mg/L (for the Enterobacterales and A. baumannii), or ≥ 4 mg/L for P. aeruginosa, an alternative agent should be considered, as bactericidal levels are unlikely to be achieved for organisms with MICs in excess of these cut-offs. Consultation with a clinical microbiologist or infectious diseases specialist is advised to assist with appropriate interpretation of colistin MICs dependent on which antimicrobial susceptibility testing committee guideline (EUCAST or CLSI) is followed by the local laboratory.

The CLSI and EUCAST are the two organisations that provide susceptibility breakpoints for colistin.

Challenges in the setting of breakpoints for colistin include: (1) difficulties in providing a reproducible MIC below 2 mg/L because of challenges in the test system; (2) in vivo levels of 2 mg/L required for bactericidal action, which is very difficult to obtain in patients with normal renal function; (3) MIC distribution data indicates that 2 mg/L is the most reliable cut-off point to separate wild type isolates from those with resistance mechanisms for the Enterobacterales and A. baumannii, but higher (4 μg/L) for P. aeruginosa.¹⁴

The current (2022) CLSI guidelines have set colistin breakpoints for Enterobacterales, P. aeruginosa, and Acinetobacter spp at ≤ 2 mg/L (intermediately susceptible) and ≥ 4 mg/L (resistant), with no susceptible category (Table 2).¹² The CLSI definition of the intermediate category implies uncertainty related to susceptibility testing, clinical outcome, dosing, and administration.

In contrast, EUCAST has maintained the susceptibility category, for A. baumannii and the Enterobacterales at ≤ 2 mg/L, and resistant if > 2 mg /L. Pseudomonas aeruginosa has slightly higher breakpoints (susceptible if ≤ 4 mg/L and resistant if > 4 mg/L) (Table 2).¹³

European Committee on Antimicrobial Susceptibility Testing introduced brackets for the colistin breakpoint. This is to warn against using colistin without additional therapeutic measures. Breakpoints in brackets represent the epidemiological cut-off value (ECOFF), which distinguishes isolates with and without acquired resistance mechanisms.¹³

Area under the concentration time curve (AUC)/minimum inhibitory concentration (MIC) (fAUC/MIC) is the recommended PK/PD parameter for colistin efficacy.

An AUC_0–24 of 50 mg.h/L is considered acceptable to achieve adequate efficacy in the presence of isolates with a MIC of ≤ 2 mg/L.² Efficacy data are lacking for the paediatric population but extrapolation from adult data suggests that a steady state concentration (C_ss) of 2 mg/L equates to an AUC_0–24 of 50 mg.h/L.² However, adult and paediatric pharmacokinetic studies have shown extensive interpatient variability in C_ss.^15,16,17

All neonates, infants and children requiring colistin should receive a colistin loading dose of 4 mg – 5 mg CBA/kg body weight (equivalent dose in million units MU is 118 000 IU/kg – 150 150 IU/kg), which should precede maintenance doses of colistin 2.5 mg CBA/kg/dose 12 hourly in children > 2 years (equivalent dose in MU is 74 000 IU/kg). We recommend the same maintenance dosing strategy in children < 2 years until further data are available.

Colistin loading doses in adult patients are recommended and well established.² A colistin loading dose prevents exposure to sub-therapeutic concentrations for a prolonged period during initial treatment, as colistin concentrations rise slowly after administration.²

The Food and Drug Administration (FDA) and the European Medicines Agency (EMA) currently suggests no paediatric loading dose with a maintenance colistin dose of 2.5 mg to 5 mg CBA/kg per day.^18,19 The reason for no loading dose is mainly because of a lack of robust PK and safety data in the paediatric population. Recent PK studies in children and neonates have however shown that the FDA- and EMA-recommended colistin doses are insufficient for optimal efficacy according to recommended Pharmacokinetics (PK), pharmacodynamics (PD) parameters.^15,20,21 To our knowledge, there is only report on the pharmacokinetics of colistin in neonates.²⁰ This was a prospective, open label study, performed in neonates (5–15 days of life) receiving colistin within a neonatal intensive care unit.²⁰ Further elaboration on loading dose is available in the supplementary document.

In addition, paediatric colistin studies reported reversible nephrotoxicity in 3% – 10% of patients,¹⁷ but the impact of immature neonatal metabolic and renal function on the metabolism and elimination of colistin has not been assessed.

Based on evidence that the standard 5 mg CBA/kg per day colistin doses without loading doses produce suboptimal exposures,^22,23 and given the emerging paediatric and neonatal PK data of colistin showing benefits of a colistin loading dose and likely subtherapeutic concentrations with the suggested maintenance dose, we recommend a colistin loading dose of 4 mg – 5 mg CBA/kg body weight, followed by 2.5 mg CBA/kg twice daily maintenance dose in neonates, children, and infants. This should be performed with close monitoring of renal function (see ‘How should adverse effects be monitored and how often’).

It is important to observe that therapeutic drug monitoring to guide colistin dosing is not available in South African at the time of publication.

Data on colistin pharmacokinetics in the paediatric population with renal impairment are lacking. Conclusive recommendations on dosage and/or dose interval in patients with renal impairment cannot be established. Where alternative treatment options are available, colistin should be avoided. Where no alternative is available, individual cases should be discussed with a clinical pharmacologist, microbiologist or paediatrician to assist with dose adjustment.

Overall, fewer paediatric patients experience colistin-associated acute kidney injury than adults.²² Colistin renal dose adjustments in adults are well established but are lacking for the paediatric population.^2,18 Most paediatric studies evaluating the pharmacokinetics of colistin exclude patients with renal impairment.^20,23,24 A recent population pharmacokinetic study of colistin explored covariates influencing target attainment in a population with a median age of 2.6 years (interquartile range [IQR] 0.8–6.8 years). Creatinine was a significant covariate and colistin dosing adjustments are proposed in impaired renal function. However, the proposed dosage changes have not been validated.

The pharmacokinetics of colistin in paediatric patients receiving renal replacement therapy (RRT) have not been adequately evaluated. No recommendations can be made concerning colistin dosing in patients receiving RRT. In adult patients, dose adjustments are recommended depending on the type of RRT received.² In the absence of data, colistin dosing regimens during RRT should be selected through a multidisciplinary team approach, including neonatologists, nephrologists, microbiologists, clinical pharmacologists, infectious disease specialists, and chemical pathologists.

Colistin may be administered as a slow bolus injection or as a slow infusion over 30 min.

The contents of the vial can be reconstituted and administered as a slow bolus injection over 3 to 5 min or as an infusion over 30 min.^18,19 The volume chosen for infusion should be determined by patient’s fluid requirements. Table 3 is a CMS conversion table.¹⁸ Refer to Online Appendix 1 for a practical, stepwise approach to administering CMS.

Combination therapy with a second active antibiotic is recommended in those patients with severe sepsis or septic shock at presentation, bloodstream infection (BSI) with a non-urinary/non-biliary source of infection or severe underlying disease. Factors for consideration when selecting the second active antibiotic (meropenem, tigecycline, aminoglycoside, other) include the site(s) of infection, the antibiotic MICs and patient renal function. High dose prolonged infusion meropenem can be used when the MIC is ≤ 8 mg/L.

For non-severe infections (e.g. uncomplicated UTI) colistin monotherapy can be considered in the absence of other appropriate treatment options.

Several factors support combination therapy: in vitro data showing synergy of antibiotic combinations, lower efficacy of colistin monotherapy compared with β-lactam monotherapy, the potential to reduce mortality in severely ill patients, lower risk of resistance development (e.g. against colistin), and shorter treatment duration. However, several facts argue against combination therapy: the possible rise in resistance rates because of an overall increase in selection pressure following greater release of antibiotics into the environment, higher rates of adverse effects (such as nephro- and ototoxicity because of colistin or aminoglycosides), increased Clostridioides difficile-associated infections, fungal infections, higher costs, and possible antagonism.²⁵

Although data supporting combination therapy is sparse, and largely for adults with Klebsiella pneumoniae Carbapenemase (KPC) BSIs, recent meta-analyses concluded that combination therapy reduces mortality and improves clinical outcomes in patients with BSIs because of CPE.^25,26

The newer β-lactam-β-lactamase inhibitor combinations are not readily accessible in many countries, resulting in limited treatment options being available for CRE infections. Available treatments include colistin, the aminoglycosides, tigecycline and PK/PD-optimised doses of meropenem. The selection of specific antibiotics must be based on the likelihood of achieving therapeutic drug levels at the site of infection (site of infection, antibiotic MIC and PK/PD parameters) without irreversible or severe adverse effects related to the drug.

Data from observational studies on combination therapy are inconsistent.² Some observational studies showed benefit of meropenem-based combination therapy for CPE (mainly KPC) when the meropenem MIC is 8 mg/L or lower.^27,28 The second agents used in these patients varied and included colistin, tigecycline and aminoglycosides. In contrast, the AIDA and OVERCOME randomised controlled trials (RCTs) did not find a significant survival benefit of colistin-meropenem combination therapy over colistin monotherapy in patients with severe CPE infections.^29,30

Other studies show a benefit of combination therapy only with severe sepsis or septic shock at presentation, BSI with non-urinary/non-biliary source of infection, or severe underlying disease.^31,32 A systematic review and meta-analysis comparing monotherapy to combination therapy for MDR Gram-negative infections demonstrated reduced mortality in BSIs and in infections caused by CPEs when combination therapy comprising two active antibiotics was used.²⁵ There was no reduction in mortality when combination therapy included only one active antibiotic. Clinical cure rates with mono- and combination therapy were the same. Overall, the quality of studies included in this analysis was low. The improved outcomes observed when colistin is combined with other antibiotics may be because of suboptimal pharmacokinetic properties of colistin.

Based on the available evidence, combination therapy is recommended in patients with CPE infections who have severe sepsis or septic shock at presentation, BSIs with a non-urinary/non-biliary source of infection or severe underlying disease.

Combination therapy with a second active antibiotic is recommended for severe sepsis/septic shock. Factors to be considered when selecting the second active antibiotic (meropenem, tigecycline, aminoglycoside etc.) include the site and source of infection, the antibiotic MICs, and patient renal function. Consideration may be given to the use of high dose extended infusion meropenem when the meropenem MIC is ≤ 8 mg/L and an alternate second active antibiotic is not available. Rifampicin is not recommended as a second antibiotic. If a second active agent is not available, colistin monotherapy is recommended. Where other antimicrobial agents with A. baumannii coverage are available (e.g. ampicillin-sulbactam, cefiderocol), these are recommended over colistin monotherapy.

For patients with non-severe infections (e.g. uncomplicated UTI, skin and soft-tissue) colistin monotherapy can be considered in the absence of other appropriate treatment options.

The choice of antibiotic treatment should be based on susceptibility testing balancing the expected clinical success rate against the risk of development of ABR and the risk of severe side effects.²⁵

The AIDA RCT compared colistin monotherapy to colistin-meropenem combination therapy in 406 patients, for severe infections (largely pneumonia and bacteraemia caused by A. baumannii), they however, did not find a difference in clinical outcomes.²⁹ A planned sub-analysis of the subset of infections caused by organisms with meropenem MICs ≤ 16 mg/L was not possible because of low numbers. The OVERCOME RCT, which also compared colistin to colistin-meropenem combination therapy for carbapenem-resistant Gram negative pneumonia and BSI found no survival benefit with combination therapy.³⁰ The XDR A. baumannii isolates prevalent currently have high carbapenem MICs (> 8 mg/L) (unpublished data) and the findings of the cited studies do not support the routine use of colistin-carbapenem combination therapy.

Similarly, despite in vitro studies demonstrating synergy when colistin is combined with rifampicin for A. baumannii, current clinical data do not support the use of colistin-rifampicin combination therapy.^33,34

The XDR A. baumannii isolates prevalent in the South African public health sector currently have high carbapenem MICs (> 8 mg/L). The lowest concentration of meropenem and imipenem inhibiting 50% and 90% (MIC₅₀ and MIC₉₀ of 2019 paediatric bloodstream A. baumannii isolates for Gauteng, KwaZulu-Natal, Free State, and Western Cape provinces are > 8 mg/L (unpublished data courtesy of Prof Olga Perovic, Antimicrobial Resistance Laboratory, National Institute for Communicable Diseases [NICD]).

The evidence supporting combination therapy for A. baumannii is of low quality. However, the pharmacokinetic properties of colistin result in suboptimal drug levels at some infection sites when administered intravenously. This, together with its narrow therapeutic window, can result in limited clinical efficacy for many infections.² In addition, delays in appropriate therapy (because of the MDR phenotype) may result in high bacterial burdens. Hence, combination therapy with a second active antibiotic is recommended in those patients with severe sepsis/septic shock. If a second active agent is not available, colistin monotherapy is recommended. For patients with non-severe infections (e.g. uncomplicated UTI) colistin monotherapy can be considered in the absence of other appropriate treatment options.

Combination therapy with a second active antibiotic is recommended for severe sepsis or septic shock at presentation, BSI with a non-urinary/non-biliary source of infection or severe underlying disease.

For patients with non-severe infections (e.g. uncomplicated UTI), colistin monotherapy can be considered in the absence of other appropriate treatment options.

There is little data available limiting a robust recommendation. In the small number of carbapenem-resistant P. aeruginosa cases included in the AIDA and OVERCOME RCTs, no survival benefit was demonstrated for colistin–carbapenem combination therapy over colistin monotherapy.^29,30 Available data are mainly small retrospective studies that used a variety of active and non-active second antibiotic agents, and it is not possible to make conclusions about the value of colistin-combination therapy.

It is suggested that combination therapy is used in those patients with severe sepsis/septic shock and colistin monotherapy is used for non-severe infections.

In authors’ view, empiric colistin therapy can be considered in critically ill patients in centres where invasive infections caused by carbapenem-resistant Gram-negative pathogens are prevalent. However, empiric colistin should be stopped as soon as possible e.g. alternative antibiotic active against isolate, clinical improvement, exclusion of invasive infection, non-infectious cause for illness identified.

For most patients, empiric colistin use is strongly discouraged. It is recommended that empiric use of colistin be approved through the hospital AMS committee on a case-by-cas e basis.

When empiric colistin treatment is being considered, it should only be initiated following consultation with a clinical microbiologist, neonatologist or infectious diseases specialist as there are frequently better alternatives available. Strict guidelines must be in place to guide the empiric use of colistin to prevent overuse of this last resort antibiotic and risk of colistin resistance (see AMS recommendations).

It is controversial as to whether colistin should ever be prescribed empirically that is without confirmed or presumptive infection with a carbapenem-resistant Gram-negative pathogen.

The possible harms related to delayed initiation of potentially life-saving treatment must be weighed up against the benefits of preserving activity for this ‘last-line’ agent against highly-resistant bacteria. Table 4 outlines a step-by-step guide for prescribing colistin empirically.

Infections with XDR Gram-negative organisms are associated with increased mortality. One possible explanation for the increased mortality seen is the likely delay in starting treatment with an appropriate empiric antimicrobial, as none of these organisms would be susceptible to the standard empiric treatment regimens used. Inappropriate empiric antibiotic therapy and delays in starting appropriate antimicrobial therapy are associated with increased mortality.^35,36,37 A recent study looking at starting empiric colistin/imipenem in patients with severe sepsis showed a significant reduction in vasopressor requirement and faster improvement in inflammatory markers when treatment was appropriate.³⁸ The choice of empiric antimicrobial therapy should be based on the local epidemiology of isolated pathogens as well as culture–site specific local antibiograms (e.g., blood culture specific antibiograms). Empiric use of colistin in units with a high burden of XDR Gram-negative bacteria (> 15% XDR) may be considered, but could lead to the overuse of this last resort antibiotic and increase the risk of colistin resistance. Therefore, strict guidelines must be in place to guide the empiric use of colistin for suspected invasive infection with XDR Gram-negative bacteria. Access to facility level data in the South African public sector is available on the NICD dashboard.³⁹

Inhaled colistin in ventilator-associated pneumonia (VAP) because of colistin-only susceptible Gram-negative bacteria in critically ill children is not routinely recommended.

However, inhaled colistin, in addition to systemic colistin, may be considered in VAP because of colistin-only susceptible Gram-negative bacteria for treatment failure by systemic colistin alone and with the availability of small particle nebulisers.

Although inhaled colistin may be a beneficial adjunct to IV colistin by leading to shorter time to bacterial eradication, significant differences in the clinical and microbiological outcomes of children with VAP have not been demonstrated.

When considering adding inhaled colistin to systemic colistin, prior consultation with a paediatric infectious disease specialist/microbiologist is recommended.

The hydrophilic structure of colistin limits its penetration into lung parenchyma.⁴¹ Nebulised colistin potentially achieves higher concentrations in the airways with less systemic toxicity, compared with intravenous (IV) colistin.^42,43 To reach the lung parenchyma, the particle size, expressed as mass median aerodynamic diameter (MMAD), should be about 3 μm.⁴⁴ Most nebulisers are designed to effectively deliver drug to the airways, not the lung parenchyma, and create aerosols of 5 μm MMAD.⁴⁴ Nebulisers able to provide sufficiently small particles (< 3 μm) to reach the lung parenchyma include jet, ultrasonic, and vibrating-mesh nebulisers.⁴⁵ In addition, the absorption rate will depend on many factors including the location of deposition (central versus peripheral) as well as the volume and mechanical properties of airway secretions.⁴³

Although inhaled colistin is widely used for cystic fibrosis (CF), the pharmacodynamics might be different in a patient without chronic lung disease.^46,47 The bacteria in patients with CF are mainly found in the mucus rather than on the epithelial surface.⁴³

The evidence for inhaled colistin in adults with VAP because of multidrug resistant (MDR) Gram-negative organisms is of low quality.⁴⁸ Expert opinion from professional societies, such as the Infectious Disease Society of America (IDSA) and the European Society for Clinical Microbiology (ESCMID) are conflicting.⁴⁹ The evidence for inhaled colistin in the paediatric population is limited to small retrospective studies, with no respiratory complications reported: most infants received inhaled colistin in addition to active systemic antibiotics.^50,51 Monotherapy with inhaled colistin (without systemically administered antibiotics) was successful in 17 neonates.^52,53 A matched case control study found that infants treated with nebulised colistin plus systemic colistin, had better outcomes than infants treated with systemic colistin alone.⁵⁴ The addition of inhaled colistin to IV colistin led to a shorter time to bacterial eradication in critically ill children with VAP because of colistin-only susceptible GNB. However, it did not lead to a significant difference in the clinical and microbiological outcomes of VAP.⁵⁵

A PK/PD study suggested that a dose of 4 mg CBA/kg (120 000 IU/kg) attained high colistin levels in tracheal aspirate from neonates for 12 h.⁵⁶

Adverse effects associated with inhaled colistin are bronchospasm and nephrotoxicity. High concentrations may cause damage to the airways. Nebulised steroids and β-2-agonists can be used to prevent and treat bronchospasm.⁴⁴

The use of intraventricular/intrathecal colistin in combination with IV colistin can be considered in patients with suitable indwelling devices and clinical/microbiological indications for colistin therapy (See above recommendations for indications). In addition, for patients without a suitable indwelling device but persistent CNS infection despite maximum recommended IV colistin dosing, neurosurgery and microbiology/infectious diseases consultation is recommended prior to initiating intraventricular/intrathecal colistin. Proposed dosing of intrathecal/intraventricular colistin is detailed in Table 5.

Colistin does not cross the blood-CSF barrier well.⁵⁷ Intrathecal (through the lumbar thecal sac) or intraventricular (lateral ventricle) dosing of colistin bypasses this barrier and can achieves much higher CSF colistin levels than with IV dosing.

Intrathecal and intraventricular colistin administration has not been well studied.⁵⁸ Available low quality evidence suggests possible limited benefit and no concerns of long-term/irreversible harm.^{57,58.59,60,61,62} We therefore, recommend consideration in patients with suitable indwelling devices, such as lumbar or external ventricular catheters/drains. For patients without a suitable indwelling device but persistent CNS infection despite maximum recommended IV colistin dosing, neurosurgery and microbiology/infectious diseases consultation is strongly recommended prior to initiating intraventricular/intrathecal colistin.

Ventricular volume and CSF drainage rates must be considered when selecting a dosing regimen.⁶³ The CSF volume in neonates is small (approximately 5 mL) compared with that in older infants (50 mL) and adults (125 mL – 150 mL).⁵⁶ Doses of 4 mg CBA/day (120 000 IU/day) in adult patients yields CSF through concentrations above 2 mg/L.^23,64 The IDSA suggests that for infants, adult intraventricular antimicrobial doses should be reduced by 60% or more. A retrospective review of colistin use for MDR Gram-negative infections in a Pakistan neonatal unit, briefly described the management of 15 cases of meningitis. Seven neonates received intraventricular colistin, five of whom also received IV colistin. The five neonates who received both intraventricular and IV colistin, and one of two who received only intraventricular therapy, survived. None of the eight neonates who received only IV colistin survived. The basis for the dose selection in this unit was not provided.⁶¹

Individualised dosing based on predicted ventricular volume and CSF drainage is recommended. Multidisciplinary (neurosurgery, microbiology, infectious diseases, paediatrics and clinical pharmacology) consultation is recommended for dose selection.

If administering intrathecal or intraventricular colistin, a single daily dose is recommended.

Definitions of acute kidney injury are detailed in Online Appendix 1. The ICU patients requiring organ support require more frequent monitoring but if not requiring organ support, monitoring every 72 h is sufficient. Ideally, monitoring should be clinician directed. Collect a specimen for baseline renal function prior to colistin initiation, but do not delay initiating colistin loading dose while awaiting results.

Adverse effects because of colistin have been reported since the early 1960s, mostly from adult studies, with rates as high as 50%.⁶⁵ Reported adverse effects included nephrotoxicity and less commonly, neurotoxicity.

Data on colistin safety in the neonatal/paediatric population is limited and mostly restricted to retrospective reviews, with recent reported rates of nephrotoxicity ranging from 0% to 24%. These rates have reduced compared with older reports, attributed to improved fluid and electrolyte management in intensive care units, as well as better monitoring of renal function and reduced concomitant use of other nephrotoxic agents. The suggested mechanism of nephrotoxicity is increased membrane permeability, causing tubulopathy, influx of electrolytes and water, leading to cell oedema and lysis.⁶⁵ Definitions used for nephrotoxicity are not standardised and include an increase in serum creatinine > 50% above baseline, a decrease in urine output below 50% of baseline or < 1 mL/kg/h, or an increase in serum creatinine of > 0.5 mg/dL (44 μmol/L).^66,67,68 A case series reported a 19% renal toxicity rate in neonates receiving colistin.⁶⁹ Another retrospective study in 104 children reported nephrotoxicity in 10.5% of patients receiving colistin.⁶⁶ These children, however, were also receiving other concomitant nephrotoxic drugs and none on colistin alone developed nephrotoxicity. Another study reported nephrotoxicity in 2 of the 18 neonates treated with colistin.⁶⁷ In other retrospective studies in neonates, including preterm neonates, colistin was well tolerated, with no reported cases of renal impairment.⁷⁰ Electrolyte imbalances, particularly hypomagnesemia, hyponatremia and hypokalemia, are reported in those on colistin. In a retrospective study 12 neonates on IV colistin, 2 had significant hyponatremia and hypokalemia. In this study magnesium replacement was required at least once for all patients.⁷¹ Another case-control study that compared 47 neonates who were given colistin with 59 neonates treated with other antimicrobial agents concluded that colistin use was significantly associated with hypokalemia and hypomagnesemia.⁷²

Measures used to prevent or limit nephrotoxicity include strict monitoring of renal function with dosage adjustments when necessary, proper hydration and limiting co-administration of other nephrotoxic drugs.⁹

Neurotoxicity is the second most common adverse effect reported with polymyxins.⁷³ Manifestations of neurotoxicity include dizziness, generalised weakness, muscle weakness, facial and peripheral paraesthesia, partial deafness, visual disturbances, vertigo, confusion, hallucinations, seizures, ataxia, neuromuscular blockade and apnoea.⁶⁵ Neurotoxicity usually develops in the first 4 days of treatment.⁷⁴ In a neonatal study, neurotoxicity was not an obvious problem. However, the authors observe that sedation and mechanical ventilation may have affected these findings.²⁰ In a systematic review of polymyxin toxicity, neurotoxicity was reported more frequently in older literature than more recently, but still old, published literature (until 2005).⁶⁵ Neurotoxicity was reported more frequently after a loading dose was compared in a single study, but this finding was not statistically significant.⁶⁵ Concomitant administration of colistin with curariform muscle relaxants and other neurotoxic agents must be avoided because these combinations may trigger progression to neuromuscular blockade.⁶⁵ Where possible, concomitant use should be avoided. Where unavoidable, close daily monitoring by physical examination is necessary and daily review of the prescriptions necessary to determine the need for concomitant use. Of note, however, detecting of neurotoxic symptoms in neonates is challenging.

Treatment duration is dependent on site of infection, its severity, source control attainment, and PK/PD. Duration of treatment may differ according to indication. To prevent development of resistance to colistin, duration of therapy should be as short as possible, and guided by clinical and biomarker responses. Duration of therapy should be discussed with infectious diseases specialist/microbiologist:

The optimal duration of antibiotic therapy depends on many factors. The integration of signs of resolution, biomarkers, clinical judgement, and microbiologic eradication might help to define this optimal duration in patients with life-threatening infections caused by XDR Gram-negative bacteria. It is important to observe that prolonged therapy is not required for infections caused by MDR and XDR pathogens compared with susceptible isolates of the same species.⁷⁵

The previous South African colistin use guideline (2016) gave recommendations for colistin prescribing and administration in adults and children but did not include AMS for colistin use in humans. Although there is plentiful guidance on AMS implementation in hospital inpatients, few guidelines address colistin stewardship specifically.^{9,77,80,81,82,83} However, many generic AMS principles remain highly relevant to AMS for colistin as listed in Table 6. Table 7 outlines recommendations for monitoring colistin use according to a ‘colistin bundle’.