Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Protocol - Intervention

Strategies to prevent kidney injury from antibiotics in people with cystic fibrosis

Authors' declarations of interest

Version published: 22 May 2018 Version history

https://doi.org/10.1002/14651858.CD013032

Editorial note

This protocol will not be progressed to the review stage as there has been no progress with the review in four years. If new authors are identified, they will take this topic forward.

Collapse all Expand all

Abstract

Objectives

This is a protocol for a Cochrane Review (intervention). The objectives are as follows:

To assess the benefits and harms of strategies (such as altering the type and dose of intravenous antibiotics, the avoidance of other nephrotoxic drugs alongside the intravenous antibiotics and the use of adjuvant medication including statins and fluids) to reduce or prevent kidney damage in people with CF which occurs as a results of intravenous antibiotic treatment.

Background

Description of the condition

Cystic fibrosis (CF) is an inherited life‐limiting condition caused by a mutation in the cystic fibrosis transmembrane regulator (CFTR) gene. The disease is inherited in an autosomal recessive pattern meaning an individual requires two faulty copies of the CFTR gene, one from each parent, to inherit the disease. In the UK around 1 in 25 people are carriers of a faulty CFTR gene (Bobadilla 2002). In 2016, the UK's Cystic Fibrosis Registry identified 10,461 people diagnosed with CF in the UK (CF Trust 2017). This equates to about 1 in 2500 live births each year (Tobias 2011).

People with CF have faulty, absent or reduced numbers of CFTR channels (Cant 2014), which results in stickier mucus in the affected individuals compared to the normal population. This mucus affects multiple organ systems including the lungs, the pancreas, the gastrointestinal tract, the reproductive tract, and the hepatobiliary system.

In the lungs, the stickier mucus is more difficult to clear by coughing or by normal mucociliary clearance. Mucus contains bacteria which can be harmful and tenacious and long‐standing or chronic infection can result. In children and adults with CF, preventing and treating respiratory infection is important in optimising respiratory outcomes. To achieve this, they receive antibiotics throughout life and some of these treatments can result in kidney damage (nephrotoxicity) (Wargo 2014). Initially, treatment for a respiratory infection is usually less frequent and antibiotic choices include less nephrotoxic narrower‐spectrum oral and intravenous antibiotics (Conway 2008). However, over time there is tendency for respiratory pathogens to become more resistant to treatment and broader‐spectrum intravenous antibiotics are required more frequently (Conway 2008). Some of these antibiotics, particularly aminoglycosides which target Pseudomonas aeruginosa (P aeruginosa), are known to be associated with nephrotoxicity (Al‐Aloul 2005). Amikacin is an aminoglycoside increasingly used for the treatment of non‐tuberculous mycobacteria (Martiniano 2017). Other antibiotics can also damage the kidneys including beta lactams, which can cause an interstitial nephritis (Perazella 2014). Colistin, another anti‐pseudomonal drug, can cause kidney damage particularly when used in conjunction with aminoglycosides (Al‐Aloul 2005). Oral and nebulised antibiotics may also affect the kidneys, but this is much less common than is the case with intravenous antibiotics (Nazareth 2013). There is also evidence that while anti‐inflammatory drugs can help to slow the rate of decline in lung function (Lands 2016), these drugs, in particular non‐steroidal anti‐inflammatory drugs (NSAIDs), can also result in kidney damage (Whelton 1999).

The eradication of P aeruginosa is attempted for initial isolates using a combination of oral, intravenous and nebulised antibiotics (Langton Hewer 2017). If eradication is unsuccessful and chronic P aeruginosa develops, people with CF are likely to be prescribed long‐term nebulised anti‐pseudomonal antibiotics and will require more frequent courses of oral and intravenous antibiotics to treat respiratory exacerbations (Douglas 2009). In some centres, individuals who have not isolated P aeruginosa are still treated with anti‐pseudomonal antibiotics during exacerbations.

Damage to the kidneys can result in what is known as an 'acute kidney injury' (AKI) which is defined as any of the following:

  • an increase in serum creatinine of at least 0.3 mg/dL within 48 hours;

  • an increase in serum creatinine to 1.5 times baseline (which is known, or presumed, to have occurred within the previous seven days); or

  • urine volume less than 0.5 mL/kg/h for six hours (KDIGO 2012).

AKI incurs significant costs, extends hospital stay and often foreshortens antibiotic treatment plans. A UK study published in 2007 estimated that the incidence risk of AKI is between 4.6 and 10.5 cases per 10, 000 people with CF per year (Bertenshaw 2007). It is possible for AKI to resolve and for the individual to regain normal kidney function; however, recent evidence suggests AKI, may initiate the development of chronic kidney disease (CKD), particularly if a person suffers multiple occurrences of AKI (Hsu 2016).

CKD is defined as abnormalities of kidney structure or function (e.g. albuminuria or glomular filtration rate (GFR) below 60 mL/min/1.73 m²), present for longer than three months, with implications for health (KDIGO 2013). It has been reported that CKD can have many negative implications on a person's life (KDIGO 2013) and approximately 25% of people with CKD have been found to have depression, which is a higher rate than the general population (Palmer 2013). CKD can progress through different stages, ranging from 1 to 5; stage 5 is the most severe and is also known as end‐stage kidney disease (ESKD) or kidney failure (KDIGO 2013). In the USA, the annual prevalence of CKD (stage 3 or over) in individuals with CF is estimated to be 2.3%, which is much higher than that seen in age‐matched controls within the general population (Quon 2011). The same study reported the overall incidence rate of stage 3 or greater CKD in adults with CF in the USA was four events per 1000 person‐years of follow‐up (Quon 2011). Both the prevalence and incidence increase with age. One study in the UK estimated that in all patient populations CKD costs the NHS GBP 1.44 billion to GBP 1.45 billion per year (Kerr 2012).

CKD may eventually progress to ESKD, defined as GFR less than 15 mL/min/1.73m². ESKD has many complications including anaemia, electrolyte disturbances, bone disorders, uraemia and hypertension. People with ESKD require a form of renal replacement therapy (RRT) to correct any electrolyte or metabolic disturbances; RRT encompasses kidney transplantation and dialysis. Individuals receiving RRT are less likely to be employed than age‐matched participants, in particular those receiving haemodialysis (Helanterä 2012), which can cause a financial burden for themselves and their family. Transplantation is preferable as it reduces the mortality risk and improves quality of life (QoL) when compared with dialysis.

Description of the intervention

Whilst kidney damage resulting from antibiotic treatment is common, it is not inevitable. We understand how various treatments work and this suggests several different strategies that might be utilised to prevent or minimise nephrotoxicity.

Aminoglycosides are antibiotics used to treat respiratory infections which are potentially harmful to the kidneys. Strategies to minimise kidney injury from aminoglycosides involve looking at duration and frequency of aminoglycoside treatment, choosing one particular aminoglycoside over another or altering the dose of aminoglycosides. Previous Cochrane Reviews have examined the clinical consequences of different duration and frequency of intravenous antibiotic regimens (Plummer 2016; Smyth 2017). There was no evidence found concerning the duration of treatment (Plummer 2016), but once‐daily dosing was found to be better for the kidneys than multiple‐daily dosing (Smyth 2017). Therefore, we will not consider the effects of frequency or duration of treatment alone.

A further strategy to minimise kidney injury may involve the omission of other nephrotoxic drugs when using intravenous antibiotics. While NSAIDs are useful in reducing inflammation in the lungs (Lands 2016), they can harm the kidneys. This damage may be exacerbated when they are prescribed alongside intravenous antibiotics and to reduce this risk the duration and frequency of NSAIDs use needs to be considered. Some oral and nebulised antibiotics can also cause kidney damage, omitting these drugs while using intravenous antibiotics may reduce the risk of kidney injury.

Finally, the use of adjuvant medications (medications added to the main or primary treatment) or therapies alongside intravenous antibiotics should be considered. Statins are primarily used to reduce cardiovascular risk in individuals with elevated blood cholesterol. However, they may also be helpful in reducing kidney injury in people with CF who are receiving intravenous antibiotics (Sidaway 2004). Adequate hydration is needed for kidneys to function correctly and avoiding dehydration might avoid kidney injury. We will consider the use of enteral and intravenous fluids alongside intravenous antibiotics to see if this reduces the risk of kidney injury.

How the intervention might work

Aminoglycosides are bactericidal antibiotics that bind to the bacterial 30S ribosomal subunit inhibiting protein synthesis which prevents growth. Aminoglycosides are eliminated from the body by a process in the kidneys called glomerular filtration, but some of the drugs are re‐absorbed back into the kidney; a process that is mediated by a receptor called megalin (Schmitz 2002). This causes an accumulation of aminoglycosides in epithelial cells in the kidney (particularly in the proximal tubule), which in turn can result in the dysfunction and death of these cells leading to kidney injury (Servais 2005; Servais 2008). People with CF often have increased renal clearance (faster clearance of drugs out of the body via the kidneys) and therefore require higher doses of drugs in order to achieve therapeutic levels (de Groot 1987). An association has been found between the cumulative lifetime dose of intravenous aminoglycosides and long‐term kidney injury (Al‐Aloul 2005). So minimising aminoglycoside use can minimise these effects on the kidneys. As described above, aminoglycosides are not the only antibiotics which may cause kidney injury; others include beta lactams and colistin, but aminoglycosides are by far the worst. Tobramycin is the aminoglycoside of choice in most centres as gentamicin is even more harmful to the kidneys (Prayle 2010). We may be able to reduce kidney injury in people with CF by identifying strategies using different types and doses of intravenous antibiotics that are less nephrotoxic. It is vital, however, that the antibiotic regimens are still as effective at clearing the respiratory infection.

NSAIDs work by blocking cyclo‐oxygenase (COX) 1 and 2 enzymes, preventing the synthesis of prostaglandins and so reducing inflammation. These drugs can cause kidney injury by two main mechanisms. The first is by causing an immunological reaction which causes acute tubulointerstitial nephritis (inflammation of the kidney tissue). The second mechanism is by inhibiting COX 1 and 2 so reducing the number of prostaglandins formed. This means the blood vessels in the kidneys do not dilate as much as normal. With reduced blood flow to the kidney, there is a higher risk of injury due to lack of oxygen and glucose, especially in those with pre‐existing volume depletion (Whelton 1999). Minimising the use of these drugs when using intravenous antibiotics could reduce the risk of kidney injury.

Some antibiotics that can be given orally for respiratory infections such as ciprofloxacin, azithromycin and beta lactams have been associated with kidney injury (Nazareth 2013). It is known that nebulised colistin can cause an AKI (Florescu 2012) and there have been cases of nebulised tobramycin damaging the kidneys (Hoffmann 2002). The omission of oral and nebulised antibiotics when using intravenous antibiotics, may reduce the risk of kidney injury.

Statins inhibit an enzyme (HMG‐CoA reductase) eventually leading to a fall in the production of cholesterol, hence their use in cardiovascular disease. It has been suggested that statins may prevent nephrotoxicity (Sidaway 2004). Inhibition of HMG‐CoA reductase results in a depletion of cellular sterols which are needed for the uptake of aminoglycosides into the kidneys. Less of the aminoglycoside is deposited in the kidneys, decreasing the death of these cells. This has been observed both in vitro (Antoine 2010) and in vivo, where animal studies have shown that this effect may be limited to certain subgroups of statins with rosuvastatin having a greater effect in murine models (McWilliam 2018).

Studies examining the use of nephrotoxic chemotherapeutic agents have shown that hydration status is important in mitigating the harmful effects of toxins on the kidney (Sharbaf 2017). Adequate hydration or hyperhydration are likely to affect the toxicity of any drug treatment. Good hydration will tend to reduce the overall concentration of drugs in the body and within the urine. Adequate hydration ensures total body water is marginally increased. Thus, the total volume in which most drugs are distributed, the volume of distribution, is slightly larger. There is a larger effect on the cells located within the kidney as any excess fluid will be excreted as urine. The concentration of all water‐soluble constituents of the blood will be lower within the glomerular filtrate and therefore there will be a lower drug concentration throughout the renal tubules.

Why it is important to do this review

With improved outcomes and longer life expectancy for people with CF, the need to protect the kidneys from long‐term damage is becoming an important consideration. Median life expectancy rose to 47 years of age in 2016 (CF Trust 2017) compared to 32.2 years in 1998 (CFF 2014). Aminoglycosides are needed to achieve the best respiratory outcomes; however, they can cause nephrotoxicity, so it is important to look at strategies to prevent kidney injury. Furthermore, increasingly people (especially adults) with CF have firm views about which intravenous antibiotics should be prescribed when these drugs are needed. This review should aid people with CF in medical decision‐making by empowering them to discuss such issues as nephrotoxicity with their clinicians and encouraging them to report symptoms like tinnitus and hearing loss.

Objectives

To assess the benefits and harms of strategies (such as altering the type and dose of intravenous antibiotics, the avoidance of other nephrotoxic drugs alongside the intravenous antibiotics and the use of adjuvant medication including statins and fluids) to reduce or prevent kidney damage in people with CF which occurs as a results of intravenous antibiotic treatment.

Methods

Criteria for considering studies for this review

Types of studies

We will include randomised controlled trials (RCTs) and quasi‐RCTs of both parallel and cross‐over design in our review. We will assess quasi‐RCTs on their merit using the Cochrane risk of bias tool and include them if we are satisfied that the groups were similar at baseline. We will assess cross‐over trials on an individual basis. If we deem the treatment to alter the condition to the extent that, on entry to subsequent phases, the participants differ from their initial state, we will exclude the trial unless we can use data from the first phase only.

Types of participants

We will include children and adults with a diagnosis of CF confirmed by genetic or sweat testing who are receiving intravenous antibiotics (of any type) for treating an infection (for any reason, e.g. planned or unplanned courses of intravenous antibiotics). Although oral and nebulised antibiotics may cause kidney injury, when used alone they are much less likely to cause nephrotoxicity than intravenous antibiotics; therefore, we will not include studies where oral or nebulised antibiotics are used alone, the participants must be receiving intravenous antibiotics either alone or in combination with other antibiotics.

Types of interventions

Firstly, we will consider the use of different types (any class) and dose of intravenous antibiotics on kidney function. This may include comparing different types of antibiotics when used alone or in combinations.

We will consider the effect of avoidance of concomitant nephrotoxic drugs alongside the intravenous antibiotics. These concomitant nephrotoxic drugs include NSAIDs, oral antibiotics and nebulised antibiotics.

We will also examine the use of adjuvant (used alongside the intravenous antibiotics) medications or therapies such as statins, enteral or intravenous fluids to reduce kidney injury.

We will compare each strategy (e.g. different antibiotic type or dose, omission of concomitant nephrotoxic medications or adjuvant drugs or therapies) with standard clinical care or placebo. Where the intervention is to omit nebulised or oral antibiotics while using intravenous antibiotics, the standard clinical care would continue the nebulised or oral antibiotics (or both) alongside the intravenous antibiotics. Where the intervention is to prescribe a different antibiotic for a P aeruginosa infection that is potentially less nephrotoxic, the standard clinical care would be to use an aminoglycoside and a beta lactam (e.g. ceftazidime) as recommended by the UK Cystic Fibrosis Trust (CF Trust 2009).

Types of outcome measures

We will include any trial where nephrotoxicity is reported as an outcome if it meets all other inclusion criteria.

Primary outcomes

  1. Nephrotoxicty (determined by the change from baseline in invasive and non‐invasive biomarkers which is indicative of kidney damage)

    1. serum (blood) creatinine levels (with thresholds as defined in each trial)

    2. creatinine clearance (using e.g. the Schwartz Estimate for children and the Cockcroft‐Gault formula for adults (with thresholds as defined in each trial))

    3. urinary excretion of protein

    4. urinary excretion of biomarkers of proximal tubular toxicity (e.g. kidney injury molecule‐1, retinal binding protein, beta‐2 microglobulin, Clara cell protein, microalbumin, N‐Acetyl‐Beta‐D‐glucosaminidase, alkaline phosphatase, alanine aminopeptidase, gamma‐glutamyl transferase or cystatin C)

    5. urine output (mL/kg/h)

Secondary outcomes

  1. Eradication of respiratory infection (defined as negative bronchoalveolar lavage, sputum or cough swab cultures at the end of treatment course)

  2. Participant‐reported symptom scores (change from baseline)

  3. Lung function parameters (change from baseline)

    1. forced expiratory volume in one second (FEV1)

    2. forced vital capacity (FVC)

    3. FEV1/FVC ratio

  4. Participant‐reported QoL scores (using a validated score e.g. CFQ‐R (Quittner 2009)) (change from baseline)

  5. Adverse effects of treatment

Search methods for identification of studies

We will search for all relevant published and unpublished trials without restrictions on language, year or publication status.

Electronic searches

The Cochrane Cystic Fibrosis and Genetic Disorders Group's Information Specialist will conduct a systematic search of the Group's Cystic Fibrosis Trials Register for relevant trials using the following terms: antibiotics.

The Cystic Fibrosis Trials Register is compiled from electronic searches of the Cochrane Central Register of Controlled Trials (CENTRAL) (updated each new issue of the Cochrane Library), weekly searches of MEDLINE, a search of Embase to 1995 and the prospective handsearching of two journals ‐ Pediatric Pulmonology and the Journal of Cystic Fibrosis. Unpublished work is identified by searching the abstract books of three major cystic fibrosis conferences: the International Cystic Fibrosis Conference; the European Cystic Fibrosis Conference and the North American Cystic Fibrosis Conference. For full details of all searching activities for the register, please see the relevant section of the Cochrane Cystic Fibrosis and Genetic Disorders Group's website.

We will search the following databases:

  • Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library (www.thecochranelibrary.com);

  • MEDLINE Ovid (1946 onwards);

  • Embase HDAS (Healthcare Databases Advanced Search) (1974 onwards);

  • Open Grey (www.opengrey.eu/).

We will search the following trials registries:

  • The World Health Organization International Clinical Trials Registry Platform (www.who.int/trialsearch);

  • US National Institutes of Health Ongoing Trial Register Clinicaltrials.gov (www.clinicaltrials.gov);

  • International Standard Randomised Controlled Trial Number (ISRCTN) Registry (www.isrctn.com).

For details of our search strategies, please see Appendix 1.

Searching other resources

We will check the bibliographies of included trials and any relevant systematic reviews identified for further references to relevant trials. We will contact experts and organisations in the field to obtain additional information on relevant trials.

Data collection and analysis

Selection of studies

Once we have the complete list of identified trials, one author (ND) will check for and remove any duplicates. Two authors (ND and WC) will then review all titles and abstracts independently and discard references which clearly do not meet the inclusion criteria. We will attempt to resolve any disagreements by discussion, but if we can not reach a decision, the third author (FG) will act as an external arbiter to mediate until we can reach a final conclusion. Once we have screened trials on the basis of title and abstract, we will obtain full copies of the remaining references and screen these, again independently, using a standardised screening form customised for this review.

We will consider trials in any language and will translate them as necessary. We will include trials published as full texts and any abstracts if they present results. If there are no results presented within the abstract or on any trial registry sites, then we will classify the trial as 'Awaiting classification' until more information is available. Similarily with ongoing trials, we will include any trial meeting our inclusion criteria.

We will present the results of the search using a standardised flow chart.

Data extraction and management

Two authors (ND and WC) will independently extract data from the included trials. We plan to collect data using the data extraction forms on Covidence, an online software program that provides detailed data extraction forms for Cochrane Reviews (Covidence 2017).

We will collect data on:

  • participant characteristics;

  • trial characteristics and trial design;

  • intervention and comparator;

  • outcome data (we will report data for each outcome separately).

One author (ND) will check the two independently completed data extraction forms for discrepancies and if there are any which we can not resolve by discussion, the third author (FG) will arbitrate.

We will enter the data into Review Manager 5 software for analysis (RevMan 2014). We plan to report data at up to one week, up to two weeks, up to one month, up to three months, up to six months and up to one year. If data are reported at other time points we will consider reporting these too.

Where more than one intervention has been undertaken in a single group then we will report the combined effectiveness.

Assessment of risk of bias in included studies

We (ND and WC) will independently use the risk of bias tool as described in the Cochrane Handbook for Systematic Reviews of Interventions to assess the risk of bias across seven domains (sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting and (optionally) 'other issues') (Higgins 2011). We will attempt to resolve any disagreements by discussion, but if we can not reach a decision, the third author (FG) will act as an external arbiter to mediate until we can reach a final conclusion.

If the trial demonstrates methods of randomisation that are correct and appropriate, e.g. random number tables or computer‐generated lists, we will rank the sequence generation domain as having a low risk of bias. Where these are inadequate, e.g. allocation based on alternation, we will rank the trial as being at high risk and where it is unclear from the description given, then we will rank it as having an unclear risk of bias.

We will also look for methods of concealment of the allocation sequence from the researchers, and if we deem these to be adequate, e.g. central allocation or sequentially numbered drug containers of identical appearances, then we will rank the trial as having a low risk of bias for this domain. Where these are inadequate, e.g. the allocation may be foreseen, we will rank the trial as being at high risk and where it is unclear from the description given, then we will rank it as having an unclear risk of bias.

Similarily for blinding, the trial should state that participants and personnel were blinded or if there was no blinding we must judge that outcomes would not have been affected by lack of blinding, in order to have a low risk of bias for this domain. If there was no blinding and the outcome is likely to be affected by the lack of blinding then we will rank the trial as having high risk of bias for this domain.

We will also look for the blinding of outcome assessors. Blinding of outcome assessment must be ensured for this domain to rank as low risk of bias. If there was no blinding but we judge that outcomes would not have been affected by lack of blinding then this can also be ranked as low risk of domain. If there was no blinding and the outcome is likely to be affected by the lack of blinding then we will rank the trial as having a high risk of bias for this domain.

For the domain of incomplete outcome data, we will extract information on missing data and how the investigators recorded participant withdrawals and loss to follow‐up. We will also look at whether missing data were equally distributed between the intervention and control groups. If we agree that missing data have been accounted for adequately, then we will judge the trial to be at a low risk of bias. We will record the trial as having a high risk of bias if the investigators do not report missing data adequately and will record it as having an unclear risk of bias if we are unable to see how the missing data have been reported. We will assess each included trial to determine whether the investigators used an intention‐to‐treat (ITT) analysis and again, once we have reached an agreement, we will rank the trials as being at a high, low or unclear risk of bias.

We will search for trial protocols to assess outcome reporting. If we can not locate the protocol, we will assess outcome reporting based on a comparison between the methods section of the full published paper and the results section. If the trial investigators report all outcomes in the paper, we will record a low risk of bias for selective reporting. If the paper states that investigators measured outcomes, but they do not report the results of these, we will rank the papers as being at high risk. If it is unclear to the us whether the trial reports all outcome measures, then we will state this and rank it as unclear for this domain.

We will look for any other potential sources of bias in the included trials and will record what we find. If we cannot find any other source of bias, then we will rank the trial as having a low risk for this domain and high risk if the opposite is true.

We will present the results of the risk of bias assessment both individually and in a summary table.

Measures of treatment effect

For continuous data (serum creatinine levels, creatinine clearance, urinary excretion of protein, urinary excretion of biomarkers of proximal tubular toxicity, urine output, lung function parameters, participant‐reported QoL scores and participant‐reported symptom scores) we plan to record the mean change and standard deviation (SD) from baseline for each group. We intend to calculate a pooled estimate of treatment effect using the mean difference (MD) and 95% confidence intervals (CIs). Where trials use different units of measurement or measurement scales for reporting the same outcome (which is likely to be true for QoL and symptom scores) we will use the standardised MD (SMD) to report the results. Where trials only report only a pre‐intervention mean (SD) and post‐intervention mean (SD) then we can calculate the mean change but not the SD of the change; we will report these results narratively.

For dichotomous data (adverse effects and eradication of respiratory infection), we will calculate a pooled estimate of the treatment effects for each outcome across trials using risk ratio (RR) and 95% CIs where appropriate.

If kidney injury is reported as a dichotomous outcome (i.e. present or not) without biomarker measurements, we will attempt to contact the author for these continuous data. However, if this is not possible or data are not available, we will assess the reported dichotomous data using the RR and corresponding 95% CIs.

Unit of analysis issues

We will assess any trials using a cross‐over design to establish which data we can include in the analysis. We will include the trial if the authors have taken account of the cross‐over design in the analysis, any carry‐over effect (i.e. included a washout period for the intervention) and within‐person differences. Where the original authors have not analysed the data appropriately, we may be able to include data from the first phase of the cross‐over trial as if it were a parallel design; although the advantage of the cross‐over design (using participants as their own controls) would be lost (Elbourne 2002).

If we find trials which are multi‐arm they will possibly fall into more than one comparison. Where the two active treatment arms are different, e.g. statins or intravenous fluid, we will analyse each treatment arm separately against placebo.

Dealing with missing data

We will attempt to contact the trial investigator(s) for additional information if there are insufficient data or if there is uncertainty about data presented in the included trials. We will undertake an ITT analysis wherever possible throughout the review. We will also assess the extent to which trial investigators have employed an ITT analysis and will report the numbers of participants who dropped out of each arm of the trial, where possible. Where data are incomplete but partially available, we will use the last available measurement.

Assessment of heterogeneity

Where trials report the same outcomes which we are able to include in a meta‐analysis, we will assess the level of heterogeneity using the I² statistic. We will look at the overlap of the CIs on the forest plots to gauge the significance of the I² value.

We will base our definitions of different levels of heterogeneity on the levels described in the Cochrane Handbook for Systematic Reviews of Interventions:

  • 0% to 40% ‐ low (might not be important);

  • 30% to 60% ‐ moderate;

  • 50% to 90% ‐ substantial; and

  • 75% to 100% ‐ considerable.

The Cochrane Handbook for Systematic Reviews of Interventions states that this is a rough guide because the importance of inconsistency depends on several factors (Deeks 2011).

Assessment of reporting biases

Where we are able to include at least 10 trials, we will generate a funnel plot to attempt to identify any publication bias in the included trials (Sterne 2011). We will also attempt to identify any selective reporting in the included publications, by comparing the trial protocols with the final papers and by careful examination of the trial publications and consideration of reporting both positive and negative effects of the intervention. Where trial protocols are not available, we will compare the outcomes reported in the results section against the methods section of the paper. We will extract information on the sponsors, sources of funding and competing interests of the authors to determine the role of external bias being introduced. To minimise publication bias, we will search trial registries and contact pharmaceutical companies for unpublished data.

Data synthesis

Where we are able to combine trials in a meta‐analysis, we will use the data from the included trials to generate forest plots using the Review Manager 5 software (RevMan 2014). We will carry out separate meta‐analyses for the different interventions (e.g. different type or dose of intravenous antibiotics, avoidance of other nephrotoxic drugs, statins, fluids) versus standard care or placebo.

We will examine the level of heterogeneity to determine which type of analysis model to use. If there is low heterogeneity (less than 40%) then we will use a fixed‐effect model and if the I² statistic is greater than 40% then we will use a random‐effects model to summarize the data. However, it is important to note that as the random‐effects model allows for heterogeneity, the CI for the pooled estimate will be wider and therefore, less precise. If heterogeneity is considerable (I² over 75%), we plan to report the results narratively as it would not be appropriate in these cases to combine results in a meta‐analysis.

Subgroup analysis and investigation of heterogeneity

If a sufficient number of trials are identified (10 or more) and if heterogeneity is a greater than 40% in the included studies, we plan to undertake the following subgroup analyses for the primary outcome (nephrotoxicity):

  1. individuals receiving planned versus unplanned antibiotic treatment;

  2. children (under 18 years old) versus adults.

To test whether subgroups were statistically significantly different from one another, we will use the test for subgroup differences available in the Review Manager 5 software (Deeks 2011).

Sensitivity analysis

Where we are able to combine trials in a meta‐analysis, we will carry out sensitivity analyses to look at the effect of the risk of bias judgements. We will look at the effect of adding in and taking out trials where there is high risk of bias. We will also attempt to examine the effect of cross‐over trials on the results by carrying out a sensitivity analysis to include and exclude them.

Summary of findings table

We will report summary of findings information, with a separate table for each treatment comparison, for our chosen outcomes.

  • Blood creatinine level

  • Creatinine clearance

  • Urinary excretion of protein

  • Urinary excretion of biomarkers of proximal tubular toxicity

  • Urine output

  • Eradication of respiratory infection (defined as negative bronchoalveolar lavage, sputum or cough swab cultures at the end of treatment course)

  • Adverse effects of treatment

  • Participant‐reported symptom scores

  • Participant‐reported QoL scores

Where no data for individual outcomes are available, for that row in the table we will identify this by stating 'data not reported'.

For each outcome we will report the illustrative risk with and without the intervention, magnitude of effect (RR or MD), numbers of trials and participants addressing each outcome and a grade of the overall quality of the body of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) with comments (Schünemann 2011).