Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Protocol - Intervention

Interventions for preventing and ameliorating cognitive deficits in adults treated with cranial irradiation

This is not the most recent version

Collapse all Expand all

Abstract

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

To assess the effectiveness of interventions for preventing or ameliorating cognitive deficits in adult patients previously treated with cranial irradiation, or during treatment with cranial irradiation.

Background

Description of the condition

Cognition refers to the mental abilities that require the high‐level processing of sensory information. Such abilities include memory, executive function, thought, sensory perception, visuo‐spatial processing, concentration, attention, intellectual function, behaviour, personality and mood (Gilroy 2000). Cognitive dysfunction (or deficit) in any of these areas can have a significant impact on a person's ability to function in day‐to‐day life, including work performance, language and communication, social interactions and independent living (Meyers 1998).

Cognitive deficits are common among patients who have received cranial irradiation (Taphoorn 2004). Over 80% of primary and metastatic brain tumour patients have self reported cognitive concerns regarding memory or concentration (Lidstone 2003; Mukand 2001). In a prospective study, cognitive functioning was assessed objectively using neuropsychological testing in patients receiving cranial irradiation for the therapeutic treatment of brain metastases. Results demonstrated cognitive deficits in the domains of learning, delayed recall and recognition six to eight weeks following radiotherapy when compared to baseline scores (Welzel 2008). Patients with lung cancer receiving prophylactic cranial irradiation demonstrated reduced cognitive functioning on subjective and objective measures 6 and 12‐month follow‐up assessments when compared to baseline scores (Gondi 2013). A randomised controlled trial (RCT) documented significant cognitive deficits four months after whole brain radiotherapy compared to patients treated with radiosurgery alone (Chang 2009).

Neurotoxic effects of cranial irradiation

Patients can receive cranial irradiation to treat primary or metastatic brain tumours, or as prophylaxis of other cancer. Radiation can be delivered to the lesion(s) using large focused doses (stereotactic radiation), as part of standard fractionated treatments, or to the whole brain (whole brain radiotherapy; WBRT). Potential risk factors for cognitive decline following brain radiation include receiving fractionated radiation doses greater than 2 Gy, total radiation dose, brain volume and location of irradiation, divided‐dose schedules and overall treatment time (Lee 2002). Other risk factors may include combined or subsequent chemotherapy use, age, with those under seven years or greater than 60 years old at higher risk, and comorbid vascular risk factors such as diabetes and hypertension (Crossen 1994; Szerlip 2011). In the identification of treatment‐related neurotoxicity it is important to distinguish symptoms from tumour progression, recurrence or metastases as discontinuation of treatment may lead to irreversible central nervous system (CNS) injury (Dietrich 2008).

The neurotoxic effects of brain radiation can be divided into acute, early‐delayed and late‐delayed radiation encephalopathy (Sheline 1980). Acute radiation encephalopathy, within two weeks of therapy, occurs as a result of disruption to the blood‐brain barrier, and associated vasogenic oedema. Corticosteroids are used at this stage, and may improve symptoms of somnolence and headache, and prevent further neurologic decline. Early‐delayed radiation encephalopathy may occur at one to six months following completion of treatment, and symptoms of short‐term memory and attentional deficits are seen alongside drowsiness and worsening of pre‐existing neurological deficits. A return to baseline is often found within 12 months (Vigliani 1996). This phase is associated with blood‐brain barrier disruption, and reversible demyelination (Taphoorn 2004). In contrast to early complications, late‐delayed radiation encephalopathy is viewed as irreversible. This complication occurs months to years following radiation therapy and manifests as white matter lesions (i.e. leukoencephalopathy) or, in more severe forms, space‐occupying necrosis with mass effect and associated neurological dysfunction (Fink 2012).

The precise relationship between initial acute changes and late/chronic radiation damage to the brain is unknown. Clinically it is characterised by progressive mental slowing and impairment in attention and memory, with less commonly gait ataxia, urinary incontinence, apathy, and pyramidal and extrapyramidal signs (Taphoorn 2003). These cognitive deficits increase in incidence and severity over time (Klein 2002), however the exact incidence is hard to distinguish due to the range of neuropsychological tests, the population and the time at which patients are followed up (Taphoorn 2004). For example, up to 90% of adult brain tumour patients who survive for more than six months following WBRT therapy develop (some form of) cognitive impairment (Crossen 1994), and in up to 5% of long‐term survivors the cognitive impairment progresses to dementia necessitating admission to a nursing home (DeAngelis 1989; Vigliani 1996). The incidence of severe cognitive deficits/late delayed radiation encephalopathy is even higher in patients with primary CNS lymphoma, reaching nearly 100% in patients older than 60 years old (Abrey 1998). Due to these adverse effects of cranial irradiation, the benefit of radiotherapy treatment for patients with a more favourable prognosis, such as with a low‐grade glioma (LGG), or as prophylactic cranial irradiation for small cell lung carcinoma, has been the subject of much debate in the past decade (Gondi 2013).

The mechanism of cranial irradiation‐induced cognitive deficit

The mechanisms by which radiation causes cognitive decline, particularly in learning and memory, have been proposed to relate to metabolic changes, white matter changes and radionecrosis, as well as changes in neuronal function, particularly synaptic plasticity, and long‐lasting damage to hippocampal neurogenesis (Greene‐Schloesser 2013). Of those, impaired white matter radiation changes and neurogenesis are the most thoroughly studied.

The primary mechanism of delayed radiation‐induced white matter changes is associated with secondary endothelial damage and microvascular ischaemic insult (Lyubimova 2004), accompanied by a reduction in the proliferative capacity of glial cells (Van Der Maazen 1993). This leads to a decrease in the volume of cerebral white matter, which is directly associated with cognitive decline (Correa 2004; Mulhern 2004; Reddick 2006). This has been confirmed in a longitudinal study that found medulloblastoma patients receiving a cranial irradiation dose of 36 Gy to show more rapid cerebral white matter volume decrease than those receiving a cranial irradiation dose of 23.4 Gy (Palmer 2002). Rarely, these white matter lesions can increase in size and may progress to frank white matter necrosis characterised by focal cavitations in the white matter within the radiated fields (Anscher 1991). Treatment of radionecrosis involves surgical excision and steroid therapy, and recent studies using bevacizumab, an angiogenesis inhibitor, have also reported high rates of clinical and radiological responses, albeit with small sample sizes (Gonzales 2007; Levin 2011; Torcuator 2009; Wang 2012).

Neurogenesis refers to self renewing cells that may produce neurons, glial cells and lineage‐restricted precursor cells throughout life, associated with normal hippocampal functioning (Zhao 2008). This was explored in a post‐mortem study in patients with medulloblastoma treated with radiotherapy 2 to 23 years prior to analysis, which found significantly lower neurogenesis than controls, matched for age and sex (Monje 2007). Therefore radiotherapy strategies that attempt to spare the crucial areas of neurogenesis may produce better cognitive outcomes, compared to WBRT (Dietrich 2008; Peiffer 2011).

Measuring cognitive deficits

Cognitive functioning has been measured objectively using a battery of validated neuropsychological tests that assess different cognitive abilities (Gehring 2009; Meyers 1998); these can include tests such as digit span (Wechsler 1981), the Hopkins Verbal Learning Test (Brandt 1991) and the Controlled Oral Word Association test (Benton 1989), to assess working memory, short‐ and long‐term memory, and verbal fluency. Cognitive function has also been determined through the use of brief mental status evaluations, such as the Mini‐Mental State Examination (MMSE). Whilst the MMSE is often shorter than neuropsychological testing, it has been associated with poor sensitivity in detecting cognitive deficits (Meyers 2003). Other studies have used subjective patient reports of cognitive concerns, such as in memory and concentration (Lidstone 2003; Mukand 2001). These are suggested to be confounded by a patient's lack of appreciation regarding their cognitive impairments, and correlations with fatigue and depression, rather than cognitive test performance (Cull 1996). An additional consistent finding from the research literature is that correlations between subjectively assessed cognitive symptoms and objectively determined cognitive functioning are quite modest, with correlation coefficients generally ranging from 0.20 to 0.30 (Klein 2002).

Differences in the time points at which cognitive functioning is measured are also present, both in pharmacological and non‐pharmacological intervention studies. One study carried out assessments at baseline, and at four weeks of modafinil or methylphenidate use (Gehring 2012), whereas another continued to follow patients at 8, 16, 24 and 52 weeks following initiation of the drug memantine (Brown 2013). In cognitive rehabilitation studies patients were assessed at baseline, at the end of a seven‐week intervention and at six months (Gehring 2009), compared to baseline, at the end of a two‐week intervention and at three months (Locke 2008). These studies also demonstrate the variations in duration of the intervention.

The variations in tools available, use of both objective and subjective measures, differences in time points at which cognitive functioning is measured and the differences in intervention duration highlight the caution that must be taken when combining and generalising results and conclusions.

Description of the intervention

This review will include all interventions that aim to:

  • prevent, or

  • ameliorate

any cognitive deficits in patients who have received therapeutic or prophylactic cranial irradiation prior to, or during, participation in the study. These may include pharmacological and non‐pharmacological (medical, psychological or behavioural) interventions for the management of cognitive deficits.

Pharmacological

We define pharmacological interventions as a drug given by any route at any therapeutic dose with the intention of preventing or treating cognitive deficits in persons who have received cranial irradiation.

Studies investigating the pharmacological prevention of cognitive impairment frequently occur in patients undergoing cranial irradiation during participation. For example, memantine (Brown 2013), used in the treatment of Alzheimer's Disease, and lithium (Khasraw 2012), used in the treatment of psychiatric disorders, have both been investigated for their neuroprotective role during irradiation.

Studies of pharmacological treatment for cognitive impairment after cranial irradiation have largely focused on psychostimulants (methylphenidate and modafinil, and donepezil). Improvements in both subjective and objective cognitive functioning were seen in one study investigating the effects of methylphenidate and modafinil in brain tumour patients, 83% of which had received cranial irradiation (Gehring 2012). Donepezil, used in the treatment of Alzheimer's Disease, has also been found to show improvements in cognitive symptoms in brain‐irradiated adults (Shaw 2006).

Non‐pharmacological

We define non‐pharmacological interventions as any non‐drug intervention given with the intention of treating or preventing cognitive deficits during or following cranial irradiation. These can include, but are not limited to, medical, psychological and behavioural interventions, as well as alternative interventions such as the use of dietary supplements.

Medical interventions include any biomedical intervention given to a person in which the intervention is not primarily investigating cancer treatment or control. For example, a study exploring the use of hyperbaric oxygen therapy in cranially irradiated brain tumour patients found improvements in nine of 31 neuropsychological tests in six of seven patients, although this did not reach significance (Hulshof 2002).

Psychological interventions may include (but are not limited to) retraining, education and compensation strategies. A randomised clinical trial investigating the use of cognitive rehabilitation in glioma patients, 61% of which had received cranial irradiation, investigated computer‐based retraining and compensatory strategies. A significant improvement in objective and subjective cognitive functioning, as well as perceived burden and mental fatigue, was found (Gehring 2009).

Behavioural interventions can include exercise, as well as behavioural modification interventions.

Dietary supplements such as Ginkgo biloba have also been investigated in irradiated brain tumour patients, with some findings of cognitive improvement, although results are confounded by a high drop‐out rate (Attia 2012).

How the intervention might work

Clinical trials have explored the prevention and treatment of cognitive deficits by targeting pharmacological, psychological or behavioural pathways, as well as other biological pathways.

Pharmacological

Pharmacological interventions may prevent cognitive deficits via their neuroprotective role during WBRT such as memantine, an N‐Methyl‐D‐aspartate receptor antagonist (Brown 2013), and lithium, found to reduce oxidative distress via the glutathione system (Machado‐Vieira 2007).

Pharmacological interventions may ameliorate cognitive deficits via their involvement in critical neurotransmitter pathways. Methylphenidate is a CNS stimulant found to have a positive effect on attention due to its action on the brain centre for attention control, the fronto‐striatal network, by increasing dopamine and noradrenaline concentrations (Volkow 2002). Another centrally acting drug is donepezil, a reversible cholinesterase inhibitor involved in inhibiting the breakdown of the neurotransmitter acetylcholine. This may have a cognitive enhancing effect in cancer patients by prolonging and improving cholinergic function, associated with learning and memory (Steinberg 2011).

Non‐pharmacological

Medical interventions have also been considered to help prevent or treat cognitive deficits. Hyperbaric oxygen therapy has been used to improve damage to the nervous system by stimulating angiogenesis, the process through which new blood vessels are formed from pre‐existing blood vessels (Gill 2004).

Psychological interventions may help prevent and improve cognitive deficits by retraining cognitive capacities such as attention and memory, or via compensation strategies such as memory aids (Gehring 2009). These interventions target the plasticity of the brain, via restoration or reorganisation of function.

Behavioural interventions, such as exercise, may also help ameliorate or prevent cognitive deficits. Exercise has been associated with increases in cerebral blood flow, increased hippocampal neurogenesis, changes in neurotransmitter release and arousal levels and brain structure, and particularly through the involvement of a Brain Derived Neurotrophic Factor (Gligoroska 2012).

Other non‐pharmacological interventions, such as those involving diet modification, may also play a role in improving cognitive functioning. The dietary supplement Ginkgo biloba has been associated with regulating signalling pathways, cellular metabolism and gene transcription (Smith 2003).

Why it is important to do this review

As anti‐cancer treatments become more effective and readily available across treatment centres, patients live longer disease‐free but with long‐term sequelae of the disease and the neurotoxic side effects of treatment, which significantly impact quality of life (Cochran 2012). Subsequently, greater emphasis is being placed on disease‐free survival and quality of life and, with the establishment of neurocognitive function as a predictor of survival (Meyers 2000) and quality of life (Mitchell 2010), cognitive functioning is an essential outcome measure. There is currently no standard policy to direct treatment, and there are no systematic reviews of preventive measures or interventions for cognitive problems specifically associated with cranial irradiation in adult cancer survivors. With even mild cognitive impairment leading to negative functional and psychiatric consequences, especially if persistent and untreated, it is important to identify ways to reduce the long‐term impact of cranial irradiation on neuropsychological function.

A Cochrane systematic review will provide a high‐standard evidence base. This will guide healthcare choices by providing an examination of the evidence to distinguish the effectiveness of interventions for cognitive deficits in patients who have received cranial irradiation.

Objectives

To assess the effectiveness of interventions for preventing or ameliorating cognitive deficits in adult patients previously treated with cranial irradiation, or during treatment with cranial irradiation.

Methods

Criteria for considering studies for this review

Types of studies

Prevention

For studies investigating the prevention of cognitive deficits, we will include any studies fulfilling the following criteria:

  • they are a randomised controlled trial (RCT) or non‐randomised controlled trial (non‐RCT), including cluster and cross‐over controlled trials;

  • they have included a control group or comparison group receiving no intervention for cognitive deficit, standard care, or are compared with a normative data control group obtained from existing controls;

  • they involve an intervention aimed at the prevention of cognitive deficit in adults who are receiving cranial irradiation during participation;

  • they include cognitive performance, as assessed by neuropsychological tests, as the primary outcome, or include cognitive performance as the secondary outcome to an alternative primary quality of life measure (e.g. fatigue, mood).

We will include studies in which cognitive functioning has been measured at baseline and following intervention at any time point.

Whilst we will include studies that investigate the neuroprotective role of an intervention during cranial irradiation, we will not include those where the intervention being investigated is cranial irradiation itself, associated with treating the tumour or improving tumour control. Such excluded studies include those on:

  • hippocampal sparing techniques;

  • techniques limiting radiation dosage to healthy tissue (e.g. intensity‐modulating radiation therapy);

  • the addition of chemotherapy agents (e.g. motexafin gadolinium).

Although these techniques can be associated with reduced or limited cognitive side effects, these techniques would best fit a separate Cochrane systematic review investigating the effect of dose of radiotherapy in causing cognitive problems.

To improve the relevance of the review, we will include non‐RCTs in our search for eligible studies, however these will not be included in the main body of evidence, or any meta‐analyses, but will form the narrative discussion in the excluded studies section. Only RCTs will form the included studies section, and are eligible to be included in any meta‐analyses.

Amelioration

For studies investigating the amelioration of cognitive deficits, we will include any studies fulfilling the following criteria:

  • they are a randomised controlled trial (RCT) or non‐randomised controlled trial (non‐RCT), including cluster and cross‐over controlled trials;

  • they have included a control group or comparison group receiving no intervention for cognitive deficit, standard care, or are compared with a normative data control group obtained from existing controls;

  • they involve an intervention for the treatment of cognitive deficit in adults who have received cranial irradiation prior to participation;

  • they include cognitive performance, as assessed by neuropsychological tests, as the primary outcome, or include cognitive performance as the secondary outcome to an alternative primary quality of life measure (e.g. fatigue, mood).

We will include studies in which cognitive functioning has been measured at baseline and following intervention at any time point.

To improve the relevance of the review, we will include non‐RCTs in our search for eligible studies, however these will not be included in the main body of evidence, or any meta‐analyses, but will form the narrative discussion in the excluded studies section. Only RCTs will form the included studies section, and are eligible to be included in any meta‐analyses.

Types of participants

Prevention

For studies investigating the prevention of cognitive deficits we will include studies which include adult patients (aged 18 years and over), who have undergone cranial irradiation (whole brain or partial brain radiation) during participation in the study, for the treatment of primary or secondary brain cancer, or prophylactic treatment for other cancers.

Since these studies refer to interventions for preventing cognitive deficits, the presence of cognitive deficits at baseline is not an inclusion criterion. However, we will only include studies where cognitive functioning has been assessed via neuropsychological testing both prior to and following the start of the intervention.

Amelioration

For studies investigating the amelioration of cognitive deficits we will include studies which include adult patients (aged 18 years and over) with impairment in at least one cognitive domain, who have previously undergone cranial irradiation (whole brain or partial brain radiation) prior to participation in the study for the treatment of primary or secondary brain cancer, or prophylactic treatment for other cancers. Participants may have received cranial irradiation during childhood, but must be an adult (aged 18 years and over) during participation in the study. Cognitive impairment will be determined prior to participation via neuropsychological testing.

We will also include studies that involve only a subset of patients who have undergone cranial irradiation in the review, if this group forms a large majority (> 80%) of the study population or has been explored via subgroup analyses.

Types of interventions

Studies that will be included may have utilised pharmacological (e.g. stimulants, or neuro‐protective agents) or medical (e.g. hyperbaric oxygen therapy) approaches, or psychological (e.g. cognitive rehabilitation) or behavioural (e.g. exercise) interventions, explicitly targeted to prevent or ameliorate radiation‐related cognitive deficits.

Pharmacological interventions

We will investigate the effectiveness of any drug given by any route for any duration, and at any therapeutic dose, with the objective of preventing or treating cognitive deficits in patients who have received, or are receiving, cranial irradiation. Such drugs are likely to include psychostimulants (e.g. methylphenidate, modafinil), and may include drugs to treat cognitive deficits in other neurological conditions (e.g. donepezil, memantine). For ethical reasons, studies involving drugs may not automatically include a placebo arm. To increase the relevance of the review we will include studies without a placebo arm if the study involves a group of participants that have been randomised to a control group of some kind (e.g. treatment as usual, another active drug or allocation to a waiting list), or that have been compared to normative control data with correlation of practice effects caused by repeated neuropsychological testing.

Non‐pharmacological interventions

For medical interventions, we will investigate any medical intervention, such as hyperbaric oxygen therapy, which aim to prevent or improve cognitive deficits in patients who have received, or are receiving, cranial irradiation.

For psychological and behavioural interventions, we will review any cognitive and/or behavioural treatment given with the intention or preventing or treating cognitive deficits in patients who have received, or are receiving, cranial irradiation; these could include, but are not limited to, retraining, education or teaching of compensation strategies, physical exercise interventions or dietary supplements.

Types of outcome measures

Primary outcomes

The primary outcome will be cognitive functioning; this may be a general or composite cognitive score or individual cognitive test scores using validated neuropsychological tests (e.g. Hopkins Verbal Learning Test, Controlled Oral Word Association test). In studies involving preventative interventions, we will determine efficacy as a statistically significant improvement in cognitive functioning, or no change/decline from baseline. In studies involving treatment interventions, we will determine efficacy as a statistically significant improvement, or no change, in cognitive functioning from baseline. To increase the relevance of the review, we will not restrict eligible reviews with respect to the time point at which cognitive functioning was measured at baseline or at follow‐up. We will note and discuss the time points at which cognitive functioning was measured.

Secondary outcomes

  • Self reported cognitive functioning via interviews or questionnaires.

  • General functioning including mood/psychiatric symptoms (e.g. Hospital Anxiety and Depression Scale), self reported fatigue (e.g. Brief Fatigue Inventory) and quality of life measurements (e.g. FACT‐Br).

  • Adverse events (e.g. nausea, skin reactions, headache).

We will note and review the secondary outcomes if recorded, but they are not eligibility criteria for this review.

Search methods for identification of studies

Electronic searches

We will search the following electronic databases for published studies and conference abstracts:

  • the Cochrane Register of Controlled Trials (CENTRAL, current issue);

  • MEDLINE (1950 to present);

  • EMBASE (1980 to present);

  • CINAHL (1982 to present);

  • PsycINFO (1974 to present).

The MEDLINE search strategy is listed in Appendix 1; for other databases we will adapt the search strategy accordingly. The search strategy will not be restricted by year of publication, language or publication type.

Searching other resources

Data collection and analysis

Selection of studies

We will use the reference management database EndNote to download all titles and abstracts retrieved by electronic searching. We will remove duplicates and two review authors (JD, KZ) will independently examine the remaining references. The review authors will not be blinded to the authors or affiliations of the studies. We will exclude studies clearly not meeting the inclusion criteria and obtain full‐text copies of potentially relevant references. Two review authors (JD, KZ) will independently assess the eligibility of retrieved papers, with disagreements resolved by discussion with a third author (KG). We will document reasons for exclusion of studies.

Data extraction and management

Data extraction

We will use the recommendations from the Cochrane Handbook for Systematic Reviews of Interventions to abstract data from included trials using a data extraction form specifically designed for this review (Higgins 2011). Two review authors (JD, KZ) will complete data abstraction independently. Differences between review authors will be resolved by discussion.

Data abstracted will include the following:

  • article details (author, year of publication, journal, country and language);

  • methodology (study design, participant recruitment method, inclusion and exclusion criteria, informed consent, ethical approval, statistical analyses);

  • population demographics (geographical location, setting, age, gender, ethnicity, total number included in trial and analyses);

  • details of participants health status (including disease status, tumour pathology, treatment details, antiepileptic medication, corticosteroid use);

  • intervention (characteristics such as drug dose, preparation and route of administration, frequency and duration, detail of providers);

  • outcomes (primary and secondary outcomes assessed, method and timing of assessments);

  • results of cognitive functioning measure (neuropsychological test performance);

  • results of other outcome measures (including self reported cognitive questionnaires, quality of life, depression, fatigue and adverse events);

  • risk of bias.

Where possible, all data extracted will be those relevant to an intention‐to‐treat (ITT) analysis, in which participants are analysed in the groups to which they are assigned.

Data management

We will use Review Manager 5.3 to collate data (RevMan 2014). For continuous outcomes (e.g. cognitive performance and quality of life measures), we will extract the final value and standard deviation, and the number of patients assessed at endpoint for each treatment arm to estimate the mean difference between treatment arms and its standard error. We will note and review the time points for outcome assessment. Where participant and study details are missing, we will note these as a potential limitation of the study.

Assessment of risk of bias in included studies

We will use the Cochrane Handbook for Systematic Reviews of Interventions 'Risk of bias' tool to assess the risk of bias in included studies (Higgins 2011), including the assessment of:

  • selection bias: random sequence generation and allocation concealment;

  • performance bias: blinding of participants, personnel (patients and treatment providers) and outcome assessors;

  • attrition bias: incomplete outcome data;

  • reporting bias: selective reporting of outcomes;

  • other possible sources of bias.

A full 'Risk of bias' item list with specific criteria for each item can be found in Appendix 2.

We will interpret and report all bias criteria as having a low, high or unclear risk of bias. We will report an unclear risk of bias when insufficient information is provided, or when uncertainty over the potential for bias is present. Two review authors (JD, KZ) will apply the risk of bias tool independently and resolve differences by discussion. We will summarise results in a 'Risk of bias' graph and 'Risk of bias' summary and interpret reviews with respect to risk of bias.

Measures of treatment effect

For continuous outcomes, we will use the mean difference (MD) and standardised mean difference (SMD) with 95% confidence interval (CI).

For dichotomous outcomes we will use the risk ratio (RR) with 95% CI.

Dealing with missing data

We will not impute missing outcome data for any outcomes.

Assessment of heterogeneity

We will assess heterogeneity between studies by a formal statistical test to indicate the significance of the heterogeneity (Deeks 2001). We will investigate and report heterogeneity according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and via visual inspection of forest plots.

Assessment of reporting biases

Two review authors (JD, KZ) will review and record reporting bias. If appropriate (when a meta‐analysis includes more than 10 trials), we will examine funnel plots to assess potential small study effects, such as publication bias.

Data synthesis

If sufficient clinically similar trials are available, we will combine data for meta‐analysis using the Cochrane Review Manager software 5.3 (RevMan 2014), as follows:

  • for continuous outcomes, we will pool MDs between treatment arms at the end of follow‐up if trials measure the outcome on the same scale and at the same primary study endpoint, otherwise we will pool SMDs;

  • we will use random‐effects models for all meta‐analyses, with a 95% CI (DerSimonian 1986);

  • for dichotomous data, we will pool RRs (RevMan 2014).

Subgroup analysis and investigation of heterogeneity

Where data permit, the review will discuss studies separately using the following categories:

  • drug dose;

  • World Health Organization (WHO) tumour grade (low‐grade/high‐grade).

Sensitivity analysis

If possible, we will consider the following factors:

  • differing study quality (high or low risk of bias);

  • different classes of agents, doses or scheduling differences.

We anticipate that additional types of sensitivity analyses will be identified during the conduct of the review.