Long‐term neurocognitive and other side effects of radiotherapy, with or without chemotherapy, for glioma
Abstract
Background
Gliomas are brain tumours arising from glial cells with an annual incidence of 4 to 11 people per 100,000. In this review we focus on gliomas with low aggressive potential in the short term, i.e. low‐grade gliomas. Most people with low‐grade gliomas are treated with surgery and may receive radiotherapy thereafter. However, there is concern about the possible long‐term effects of radiotherapy, especially on neurocognitive functioning.
Objectives
To evaluate the long‐term neurocognitive and other side effects of radiotherapy (with or without chemotherapy) compared with no radiotherapy, or different types of radiotherapy, among people with glioma (where 'long‐term' is defined as at least two years after diagnosis); and to write a brief economic commentary.
Search methods
We searched the following databases on 16 February 2018 and updated the search on 14 November 2018: Cochrane Central Register of Controlled Trials (CENTRAL; 2018, Issue 11) in the Cochrane Library; MEDLINE via Ovid; and Embase via Ovid. We also searched clinical trial registries and relevant conference proceedings from 2014 to 2018 to identify ongoing and unpublished studies.
Selection criteria
Randomised and non‐randomised trials, and controlled before‐and‐after studies (CBAS). Participants were aged 16 years and older with cerebral glioma other than glioblastoma. We included studies where patients in at least one treatment arm received radiotherapy, with or without chemotherapy, and where neurocognitive outcomes were assessed two or more years after treatment.
Data collection and analysis
Two review authors independently extracted data and assessed risk of bias. We assessed the certainty of findings using the GRADE approach.
Main results
The review includes nine studies: seven studies were of low‐grade glioma and two were of grade 3 glioma. Altogether 2406 participants were involved but there was high sample attrition and outcome data were available for a minority of people at final study assessments. In seven of the nine studies, participants were recruited to randomised controlled trials (RCTs) in which longer‐term follow‐up was undertaken in a subset of people that had survived without disease progression. There was moderate to high risk of bias in studies due to lack of blinding and high attrition, and in two observational studies there was high risk of selection bias. Paucity of data and risk of bias meant that evidence was of low to very low certainty. We were unable to combine results in meta‐analysis due to diversity in interventions and outcomes.
The studies examined the following five comparisons.
Radiotherapy versus no adjuvant treatment
Two observational studies contributed data. At the 12‐year follow‐up in one study, the risk of cognitive impairment (defined as cognitive disability deficits in at least five of 18 neuropsychological tests) was greater in the radiotherapy group (risk ratio (RR) 1.95, 95% confidence interval (CI) 1.02 to 3.71; n = 65); at five to six years the difference between groups did not reach statistical significance (RR 1.38, 95% CI 0.92 to 2.06; n = 195). In the other study, one subject in the radiotherapy group had cognitive impairment (defined as significant deterioration in eight of 12 neuropsychological tests) at two years compared with none in the control group (very low certainty evidence).
With regard to neurocognitive scores, in one study the radiotherapy group was reported to have had significantly worse mean scores on some tests compared with no radiotherapy; however, the raw data were only given for significant findings. In the second study, there were no clear differences in any of the various cognitive outcomes at two years (n = 31) and four years (n = 15) (very low certainty evidence).
Radiotherapy versus chemotherapy
One RCT contributed data on cognitive impairment at up to three years with no clear difference between arms (RR 1.43, 95% CI 0.36 to 5.70, n = 117) (low‐certainty evidence).
High‐dose radiotherapy versus low‐dose radiotherapy
Only one of two studies reporting this comparison contributed data, and at two and five years there were no clear differences between high‐ and low‐dose radiotherapy arms (very low certainty evidence).
Conventional radiotherapy versus stereotactic conformal radiotherapy
One study involving younger people contributed limited data from the subgroup aged 16 to 25 years. The numbers of participants with neurocognitive impairment at five years after treatment were two out of 12 in the conventional arm versus none out of 11 in the stereotactic conformal radiotherapy arm (RR 4.62, 95% CI 0.25 to 86.72; n = 23; low‐certainty evidence).
Chemoradiotherapy versus radiotherapy
Two RCTs tested for cognitive impairment. One defined cognitive impairment as a decline of more than 3 points in MMSE score compared with baseline and reported data from 2‐year (110 participants), 3‐year (91 participants), and 5‐year (57 participants) follow‐up with no clear difference between the two arms at any time point. A second study did not report raw data but measured MMSE scores over five years in 126 participants at two years, 110 at three years, 69 at four years and 53 at five years. Authors concluded that there was no difference in MMSE scores between the two study arms (P = 0.4752) (low‐certainty evidence).
Two RCTs reported quality of life (QoL) outcomes for this comparison. One reported no differences in Brain‐QoL scores between study arms over a 5‐year follow‐up period (P = 0.2767; no raw data were given and denominators were not stated). The other trial reported that the long‐term results of health‐related QoL showed no difference between the arms but did not give the raw data for overall HRQoL scores (low‐certainty evidence).
We found no comparative data on endocrine dysfunction; we planned to develop a brief economic commentary but found no relevant economic studies for inclusion.
Authors' conclusions
Radiotherapy for gliomas with a good prognosis may increase the risk of neurocognitive side effects in the long term; however the magnitude of the risk is uncertain. Evidence on long‐term neurocognitive side effects associated with chemoradiotherapy is also uncertain. Neurocognitive assessment should be an integral part of long‐term follow‐up in trials involving radiotherapy for lower‐grade gliomas to improve the certainty of evidence regarding long‐term neurocognitive effects. Such trials should also assess other potential long‐term effects, including endocrine dysfunction, and evaluate costs and cost effectiveness.
PICOs
Plain language summary
Long‐term effects of radiotherapy for glioma treatment on brain functioning
Background
Gliomas are brain tumours that can be very aggressive and result in death within months; however, people with less aggressive gliomas (low‐grade gliomas) can survive for a number of years. Most people are treated with surgery and may also receive radiotherapy with or without chemotherapy. However, radiotherapy can damage healthy brain tissue, and we do not know enough about the possible long‐term effects of radiotherapy on brain functioning, such as memory, communication, concentration and speed of thinking (called neurocognition). Progression of the tumour can also cause deterioration in brain functioning. In this review we looked at the possible long‐term effects of radiotherapy on the brain in adults with less aggressive gliomas who had survived for at least two years after receiving treatment.
Methods and results
We searched for relevant research studies up to 14 November 2018. We only included studies with a control group (i.e. studies that included groups of people that had or had not received radiotherapy or had received different types or doses of radiotherapy). The review includes nine research studies that collected information on long‐term neurocognitive or quality of life outcomes, mostly among people with low‐grade gliomas. Altogether 2406 participants were involved in these studies. The studies looked at five different comparisons including radiotherapy versus no radiotherapy, radiotherapy versus chemotherapy, high‐ versus low‐dose radiotherapy, different types of radiotherapy, and radiotherapy versus chemoradiotherapy. Some evidence suggested that radiotherapy might increase the risk of cognitive impairment compared with no radiotherapy after surgery; however, this and evidence for the other comparisons was not convincing. This was partly because many of the people were not followed up, either because they had died or their disease had progressed, and so the resulting evidence was weak.
No studies compared effects of radiotherapy on relevant hormone functioning; we planned to develop a brief economic commentary to summarise information on whether the interventions represented a good use of health services but found no relevant studies.
Conclusions
The risk of long‐term deterioration in brain functioning associated with radiotherapy for the treatment of less aggressive gliomas remains uncertain. Further research on glioma treatment options should assess potential long‐term cognitive and hormonal side effects, costs and value for money.
Authors' conclusions
Summary of findings
| Long‐term neurocognitive and other side effects of radiotherapy, with or without chemotherapy, for glioma | ||||
| Patient or population: people with glioma surviving at least two years Settings: tertiary care | ||||
| Comparison and Outcomes | Relative effect | No of participants andstudies | Quality of the evidence | Comments |
| Intervention: radiotherapy Comparison: no adjuvant treatment Outcome: neurocognitive impairment at 5‐ to 6‐year follow‐up | RR 1.38 (0.92 to 2.06) | 1 study with data for 195 participants | ⊕⊝⊝⊝ | Outcome defined as cognitive disability deficits in at least 5 of 18 neuropsychological tests |
| Intervention: radiotherapy Comparison: no adjuvant treatment Outcome: neurocognitive impairment at 12 year follow‐up | RR 1.95 (1.02 to 3.71) | 1 study with data for 65 participants | ⊕⊝⊝⊝ | Outcome defined as cognitive disability deficits in at least 5 of 18 neuropsychological tests |
| Intervention: radiotherapy Comparison: no adjuvant treatment Outcome: neurocognitive impairment at 2 year follow‐up | RR 2.50 (0.11 to 56.98) | 1 study with data for 31 participants | ⊕⊝⊝⊝ | There was a single event for this outcome in this observational study. The outcome was defined as a significant deterioration (≥ 1 SD) in 8 out of 12 neuropsychological tests |
| Intervention: radiotherapy Comparison: chemotherapy Outcome: neurocognitive impairment at 3 year follow‐up | RR 1.43 (0.36 to 5.70) | 1 study with data for 117 participants | ⊕⊕⊝⊝ | Outcome defined as a MMSE score of 26 or less |
| Intervention: high‐dose radiotherapy Comparison: low‐dose radiotherapy Outcome: neurocognitive impairment at 2 years after treatment | RR 0.53 (0.06, 4.85) | 1 study with data for 65 participants | ⊕⊝⊝⊝ | Outcome defined as decrease in MMSE score from baseline (more than 3 points).There was serious and uneven attrition between groups in this study. |
| Intervention: high‐dose radiotherapy Comparison: low‐dose radiotherapy Outcome: neurocognitive impairment at 5 years after treatment | RR 0.16 (0.01 to3.20) | 1 study with data for 38 participants | ⊕⊝⊝⊝ | Outcome defined as decrease in MMSE score from baseline (more than 3 points). There was serious and uneven attrition between groups in this study. |
| Intervention: chemoradiotherapy Comparison: radiotherapy Outcome: neurocognitive impairment at 3 years after treatment | RR 0.37 (0.02 to 8.88) | 1 study with data for 91 participants | ⊕⊕⊝⊝ | Outcome defined as a decline (of more than 3 points in MMSE score) in cognitive state compared with baseline |
| Intervention: stereotactic conformal radiotherapy Comparison: radiotherapy Outcome: neurocognitive impairment at 5 years after treatment | RR 4.62 (95% CI 0.25 to 86.72) | 1 study with data for 23 participants | ⊕⊕⊝⊝ | Outcome defined as a decline (of more than 3 points in MMSE score) in cognitive state compared with baseline. There was serious sample attrition at 5 years. |
| GRADE Working Group grades of evidence | ||||
| Abbreviations: SD = standard deviation; MMSE = Mini Mental State Exam 1. Single study contributing data had very serious study design limitations (−2) | ||||
Background
Description of the condition
Primary brain and other central nervous system (CNS) tumours are less common than many other cancers, accounting for around 1.9% of new cancer diagnoses annually; however, they are associated with a relatively higher proportion of cancer deaths annually (2.3%), amounting to approximately 189,382 deaths worldwide in 2012 (GLOBOCAN 2012). Gliomas are brain tumours that arise from glial cells, usually oligodendrocytes and astrocytes. They occur at an annual incidence of four to 11 people per 100,000 and are more frequent in high‐income, industrialised countries (Ohgaki 2009). Gliomas are graded 1 to 4 by the World Health Organization (WHO) according to their aggressive potential in the short term. The 2007 WHO classification system (Louis 2007), used in completed clinical trials since 2007, graded gliomas based on histological characteristics only. However, in the 2016 WHO classification system, to be used in future trials, grading depends on both histological and molecular features, e.g. isocitrate dehydrogenase (IDH) status, chromosome 1p 19q, and other genetic parameters (Louis 2016). Using the 2007 WHO classification, gliomas graded 1 and 2 have low aggressive potential and are referred to as low‐grade gliomas; these include pilocytic astrocytomas (grade 1), diffuse astrocytomas, oligodendrogliomas and mixed oligoastrocytomas (grade 2). High‐grade gliomas have faster local growth rates and include anaplastic astrocytomas, anaplastic oligodendrogliomas (grade 3) and glioblastomas (grade 4). Grades correspond with prognosis: grade 1 has a good prognosis and can often be cured with surgery alone, whereas grade 4 has a poor prognosis, and can be rapidly fatal (Louis 2007). Thus, tumour grade is a key factor in deciding how to treat gliomas, particularly the need for additional treatment in the form of radiotherapy or chemotherapy or both (chemoradiotherapy) after surgery.
Description of the intervention
Most people with glioma first undergo surgery to resect (cut out) or biopsy the tumour. The latter is usually performed when resection is not possible, either due to the diffuse, infiltrative nature of the tumour, or its location near important structures. Additional radiotherapy targeting the tumour area (focal radiotherapy) is usually given immediately after surgery for high‐grade gliomas, whereas for grade 2 gliomas it can either be given immediately, or postponed if the tumour has been resected until the development of new symptoms or tumour progression (Sarmiento 2015). Fifty per cent of people with grade 2 and grade 3 gliomas survive at least seven years and four and a half years, respectively, after treatment (Buckner 2016; Cairncross 2013). However, for certain grade 2 and 3 gliomas with particular molecular features, median survival can be extended by a further seven years by the addition of adjuvant chemotherapy to radiotherapy (Buckner 2016; Cairncross 2013). Among people with grade 4 gliomas that are treated with chemoradiotherapy, only approximately 25% are alive two years after diagnosis (Stupp 2005).
Potential side effects
The treatment of glioma can be complicated by long‐term side effects that present months or years after treatment. This is due to the exposure of healthy brain tissue to radiation, which adversely affects brain plasticity (the ability of the brain to modify its connections and rewire) and repair processes (Dhermain 2016). As the frequency of side effects increases with time, these tend to be problematic for people with less aggressive tumours who survive long term, and are especially common among survivors of childhood brain tumours (Grill 1999; Seaver 1994; Spiegler 2004; Williams 2018). Certain parts of the brain such as the hippocampus, fornix and corpus callosum are more sensitive to irradiation (Connor 2017; Gondi 2012; Gondi 2018; Peiffer 2013); impairment of memory, communication, concentration and problem‐solving (neurocognition) can result. Studies among adults with low‐grade glioma show that the risk of neurocognitive impairment is increased when radiotherapy is administered to the whole brain (Gregor 1996; Surma‐aho 2001), but is less likely when radiotherapy is administered to the tumour area only (Brown 2003; Laack 2005; Taphoorn 1994; Vigliani 1996). Factors that are important to the risk of long‐term side effects in glioma treatment are the site of the tumour, the volume of brain tissue irradiated, the radiotherapy fraction size and the total radiotherapy dose. The use of chemotherapy with radiotherapy might plausibly add to the risk.
Endocrine (hormonal) dysfunction affecting adrenal (stress response) hormones, gonadal (sex) hormones, and thyroid hormones can also occur due to radiotherapy damage to the hypothalamic‐pituitary axis (Taphoorn 1995), the system that communicates with hormone‐producing glands in the body. Pituitary dysfunction is commonly diagnosed amongst children who have undergone radiotherapy for glioma, which in children frequently leads to hypothyroidism, growth hormone deficiency, and precocious puberty (Terashima 2013). In adults, recent studies suggest that pituitary dysfunction following radiotherapy for brain tumours is very underdiagnosed and that regular endocrine surveillance should be performed above a dose threshold of 30 Gy (Kyriakakis 2016; Kyriakakis 2019). In addition, fatigue, disturbed sleep and depression are also commonly reported side effects of treatment (Armstrong 2017). Such side effects can seriously interfere with a person's ability to work, maintain relationships, perform daily activities, and enjoy life (Armstrong 2016).
Why it is important to do this review
Long‐term cognitive side effects of radiotherapy were identified among the top 10 priority research questions in neuro‐oncology by the James Lind Alliance and the National Cancer Research Institute (NCRI) (JLA 2015). This is because uncertainty exists about the long‐term side effects of radiotherapy for brain tumours, especially among people with a good prognosis. Evaluating the long‐term consequences of treatment is important to understand what the real impact of this condition and its treatment are for individuals and health systems. We undertook this review to help inform clinical decision making in the context of a trend towards more aggressive early treatments for low‐grade gliomas.
The costs of care can be 'direct costs' due to health care resources used to treat the condition, or 'indirect costs' that are borne by the patient and their families. Radiotherapy is one of the highest direct costs of glioma management (Blomqvist 2000; Raizer 2015). The cost of malignant gliomas has been estimated to range between USD 50,600 and USD 92,700 (2015) per patient per year (Raizer 2015). It is, therefore, also important to understand the long‐term consequences of different glioma management strategies so that the costs and consequences of such strategies can be fully evaluated.
Objectives
To evaluate the long‐term neurocognitive and other side effects of radiotherapy (with or without chemotherapy) compared with no radiotherapy, or different types of radiotherapy, among people with glioma (where 'long‐term' is defined as at least two years after diagnosis); and to write a brief economic commentary.
Methods
Criteria for considering studies for this review
Types of studies
Randomised and non‐randomised trials, and controlled before‐and‐after studies (CBAS). We considered non‐randomised trials and CBAS for inclusion if there were no primary outcome data from randomised trials for a particular treatment comparison. We excluded cross‐over designs, case‐control studies, and studies that did not have a control group.
Types of participants
People aged 16 years of age and older with a histopathologically confirmed diagnosis of cerebral glioma who are alive at least two years after diagnosis.
In this review, as we considered late effects to be those that are present at two years or more after diagnosis among people who have a good long‐term prognosis, rather than in those that have a short‐term prognosis, we excluded studies only involving people with glioblastoma. In studies with mixed high‐grade glioma participants (grade 3 and grade 4 gliomas) we planned to extract data for the participants with grade 3 glioma only where possible.
Types of interventions
Treatment interventions after surgery (biopsy or resection of the tumour) could include the following.
-
Radiotherapy compared with no radiotherapy, which includes the following comparison subgroups.
-
Radiotherapy versus no adjuvant treatment.
-
Chemoradiotherapy versus no adjuvant treatment.
-
Radiotherapy versus chemotherapy.
-
Chemoradiotherapy versus chemotherapy.
-
-
High‐dose radiotherapy versus low‐dose radiotherapy.
-
Conventional radiotherapy versus conformal radiotherapy
-
Chemoradiotherapy versus radiotherapy.
Types of outcome measures
Studies had to report at least one of the primary outcomes in both the intervention and control groups at least two years after receiving the intervention.
Primary outcomes
-
Cognitive impairment (objective or subjective), as measured by an overall cognitive function score, a change over time score, or as a categorical outcome. This includes evaluation of cognitive impairment as individual cognitive function domains, e.g. verbal fluency, processing speed, memory, attention, and executive functioning, using a standardised measurement tool, e.g. Mini Mental State Exam (MMSE), Cognitive Failures Questionnaire (CFQ).
-
Quality of life (QoL), as measured using a standardised questionnaire, e.g. the European Organisation for Research and Treatment of Cancer (EORTC) QLQ‐C30 or QLQ‐BN20 (specific for brain cancer), or the Functional Assessment of Cancer Therapy scale (FACT‐G (general) or FACT‐Br (specific for brain cancer)).
Secondary outcomes
-
Functional impairment or disability, as measured by an overall ability score, or as a change of ability over time score, or both, using a standardised measurement tool, e.g. Karnofsky Performance Status Scale, Neurological Functions Score; or as a categorical outcome, as defined by investigators.
-
Endocrine dysfunction, as determined by use of hormonal treatment, or as defined by study investigators, or both.
-
Depression, as measured by a standardised scale, e.g. Hospital Anxiety and Depression Scale (HADS).
-
Anxiety, as measured by a standardised scale, e.g. HADS.
-
Fatigue, according to Common Terminology Criteria for Adverse Events (CTCAE), or as defined by investigators.
-
Sleep disturbances, as defined by investigators.
-
Imaging evidence of physical deficit, e.g. general brain atrophy, white matter changes, radionecrosis, stroke.
-
Social outcomes (e.g. carer strain, relationship status, employment status).
-
Second cancers.
We present evidence regarding cost of care as a brief economic commentary.
Search methods for identification of studies
Electronic searches
We searched the following databases on the 16 February 2018 and updated the search on 14 November 2018.
-
Cochrane Central Register of Controlled Trials (CENTRAL; 2018, Issue 11), in the Cochrane Library;
-
MEDLINE via Ovid (1946 to October week 5 2018);
-
Embase via Ovid (1980 to 2018 week 46).
Please refer to Appendix 1 for CENTRAL, MEDLINE and Embase search strategies.
We did not apply language restrictions to any of the searches.
Searching other resources
We searched the following for ongoing trials.
-
International Clinical Trials Registry Platform (ICTRP) (apps.who.int/trialsearch).
Where we identified through these searches ongoing trials that had not been published, we approached the principal investigators to ask for an update on the trial status and relevant data. We used the related articles feature of PubMed and handsearched the reference lists of included studies to identify newly published articles and additional studies of relevance. We also handsearched conference proceedings from 2014 to 2018 (5 years) of conferences of the British Neuro‐Oncology Society, the Society for Neuro‐Oncology, the European Association of Neuro‐Oncology and the World Federation of Neuro‐Oncology Societies for relevant ongoing or unpublished studies.
Data collection and analysis
Selection of studies
The Information Specialist at the Cochrane Gynaecological, Neuro‐oncology and Orphan Cancer Group (GNOC) downloaded all titles and abstracts retrieved by electronic searching to Endnote and removed duplicates and those studies that clearly did not meet the inclusion criteria. Review authors in teams of two (TL and RG; JE and DG) independently screened the remaining records and excluded studies that clearly did not meet the inclusion criteria. We obtained copies of the full texts of potentially eligible references and at least two review authors independently assessed these for eligibility (TL and RG, JE or DG). The two review authors concerned resolved disagreements by discussion and, if necessary, consulted the other review authors. We used Covidence to facilitate this study selection process (Covidence 2018), and document reasons for exclusion in Characteristics of excluded studies.
Data extraction and management
Two review authors (TL, TD, RG, JE or DG) independently extracted the following data from included studies to a pre‐designed data extraction form.
-
Author contact details
-
Country
-
Setting
-
Dates of participant accrual
-
Trial registration number/identification
-
Funding source
-
Participant inclusion and exclusion criteria
-
Study design and methodology
-
Study population and baseline characteristics
-
Number of participants enrolled/analysed
-
Age
-
Gender
-
Tumour grade/type
-
Type of surgery (biopsy or resection)
-
Other medication, e.g. anti‐epileptics and anti‐depressants (selective serotonin reuptake inhibitors (SSRIs))
-
-
Intervention details
-
Type of intervention
-
Type of comparator
-
-
Duration of follow‐up
-
Primary outcome/s of the study
-
Review outcomes
-
For dichotomous outcomes, we extracted the number of participants in each treatment arm who experienced the outcome of interest and the number of participants assessed
-
For continuous outcomes, we extracted the value and standard deviation of the outcome of interest and the number of participants assessed at the relevant time point in each group. We also extracted change‐from‐baseline score data where reported and noted the type of scale used
-
We extracted adjusted statistics where reported
-
Where possible, all data we extracted were those relevant to an intention‐to‐treat analysis, in which participants were analysed in the groups to which they were assigned
-
We resolved differences between reviewers by discussion or by appeal to the other review authors when necessary
-
-
Risk of study bias (see below)
Assessment of risk of bias in included studies
For randomised trials, we assessed the risk of bias using Cochrane's tool and the criteria specified in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). This includes assessment of:
-
random sequence generation;
-
allocation concealment;
-
blinding of participants and healthcare providers;
-
blinding of outcome assessors;
-
incomplete outcome data (more than 20% missing data considered high risk);
-
selective reporting of outcomes;
-
other possible sources of bias, e.g. lack of a power calculation, baseline differences in group characteristics.
For non‐randomised studies (non‐randomised trials and CBAS), we assessed the risk of bias in accordance with four criteria concerning sample selection comparability of treatment groups, namely:
-
relevant details of criteria for assignment of people with the condition to treatments;
-
representative group of people with the condition who received the experimental intervention;
-
representative group of people with the condition who received the comparison intervention;
-
baseline differences between groups controlled for, in particular with reference to age, gender, type and grade of glioma and surgical treatment.
At least two review authors (TL and at least one other) assessed risk of bias independently and resolved differences by discussion or by appeal to a third review author. We summarised judgements in 'Risk of bias' tables along with the characteristics of the included studies. We interpreted results in light of the 'Risk of bias' assessment. For more details about the assessment of risk of bias, see Appendix 2.
Measures of treatment effect
-
For dichotomous outcomes, we calculated the effect size as a risk ratio (RR) with its 95% confidence interval (CI).
-
For continuous outcomes (e.g. QoL scores) in which different measurement scales had been used, we did not pool data because time points, scales and measurement scales were too dissimilar to produce clinically meaningful estimates of effect.
Unit of analysis issues
At least two review authors (TL, TD) reviewed unit‐of‐analysis issues, as described in Higgins 2011, for each included study. These included reports where there were multiple observations for the same outcome, e.g. repeated measurements with different scales, or outcomes measured at different time points to those stipulated in the review protocol. Because data were sparse, after discussion amongst the authors we agreed to include data from different scales and time points and report the findings narratively.
Dealing with missing data
We did not impute missing data. In the event of missing data, we wrote to study authors to request the data and described in the Characteristics of included studies tables how we obtained any missing data. Where substantial volumes of data were missing, we took this into consideration in our grading of the evidence (see Data synthesis).
Assessment of heterogeneity
We did not pool data and assessed heterogeneity between studies by visual inspection of forest plots, where this was meaningful (Higgins 2003). As no data were pooled, we did not use a formal statistical test of the significance of the heterogeneity (Deeks 2001). Where there was evidence of substantial heterogeneity on visual inspection of the forest plots, we investigated and reported the possible reasons for this.
Assessment of reporting biases
Due to few included studies and limited data, it was not possible to use funnel plots to investigate reporting biases.
Data synthesis
We did not conduct meta‐analyses because data were sparse and comparisons and time points and measurements were too dissimilar for pooled estimates to be clinically meaningful. However, to help visualise the data and facilitate narrative syntheses, we created forest plots for the primary outcomes using Review Manager 5 (RevMan 5) (Review Manager 2014). For future meta‐analyses, we will use the random‐effects model with inverse variance weighting. If any trials contributing to a meta‐analysis have multiple intervention groups, we will divide the 'shared' comparison group into the number of treatment groups and comparisons between each treatment group and treat the split comparison group as independent comparisons. We will perform a meta‐analysis of the results assuming that we find at least two included studies that are sufficiently similar for the findings to be clinically meaningful.
'Summary of findings' table and reporting of results
Based on the methods described in Chapter 11 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), we prepared summary of findings Table for the main comparison to present the results of the primary outcomes, namely:
-
cognitive impairment at ≥ 2 years.
-
quality of life (QoL) score at ≥ 2 years.
We used the GRADE system to rank the quality of the evidence (Schünemann 2011). Two review authors independently graded the evidence. We resolved differences by discussion and, if necessary, by involving a third review author. Where the evidence was based on single studies, or where there was no evidence on a specific outcome, we included the outcome in the 'Summary of findings' table and graded or explained accordingly. In addition, we provided a rationale for each judgement of assumed risk in the table footnotes. In the absence of a single estimate of effect (when meta‐analysis was not possible), we rated the certainty of the effect using the GRADE approach (Murad 2017). We interpreted the results of the graded evidence based on Cochrane Effective Practice and Organisation of Care guidance (EPOC 2017).
Brief economic commentary
A brief economic commentary was planned to summarise the availability and principal findings of the economic evaluations relevant to this review. This included evaluations alongside trials and model based evaluations. The work was performed in line with current guidelines, including a supplementary search to identify economic studies (Shemilt 2018).
Subgroup analysis and investigation of heterogeneity
For the comparison 'radiotherapy versus no radiotherapy', we subgrouped studies according to the type of control group. However, as we did not pool the data we were unable to use formal tests for subgroup differences to determine whether the effect of interventions differ according to these subgroups.
Sensitivity analysis
In this version of the review, we have not performed sensitivity analysis because data were sparse. In future versions of this review, when more data are available, we plan to perform sensitivity analyses to investigate substantial heterogeneity identified in meta‐analyses of the primary outcomes, and also to estimate the effect after excluding studies at high risk of bias, to investigate how study quality affects the certainty of findings.
Results
Description of studies
Results of the search
Initial database searches conducted on the 16 February 2018 yielded the following results:
-
CENTRAL Issue 2 2018 ‒ 621 references
-
Medline: 1946 to February week 2 2018 ‒ 2302 references
-
Embase: 1980 to 2018 week 07 ‒ 2547 references
After de‐duplication and filtering out clearly irrelevant papers (e.g. studies of other cancers), we screened a total of 3197 references (including 10 references identified using the PubMed related‐articles feature) and short‐listed 57 references for full‐text screening. After full‐text screening, we classified 19 references (related to 9 studies) as included, 37 as excluded, and one as ongoing (Figure 1).
Study flow diagram (date of search 16/02/18).
The top‐up search in November 2018 yielded the following:
-
CENTRAL Issue 11, 2018 ‒ 95 additional references
-
Medline: February 2018 to October week 5 2018 ‒ 95 references
-
Embase: February 2018 to 2018 week 46 ‒ 126 references
We identified one additional study by searching the abstracts from conference proceedings. After de‐duplication, we screened 283 additional records plus the one conference abstract on title and abstract. This led to our assessment of eight full texts, seven of which we excluded and one (the conference abstract) we added to 'Characteristics of ongoing studies' (Figure 2). We identified four other potentially eligible ongoing studies by searching the clinical trials registries (NCT00457210; NCT02544178; NCT03055364; NCT03180502); and we identified one in the initial database search (CATNON 2017). We subsequently identified two related publications and added them to the related previously included studies.
Study flow diagram (date of search 9/10/18).
Included studies
The review includes nine studies that collected data on long‐term neurocognitive or quality of life outcomes: seven were conducted among people with low‐grade gliomas (Brown 2003; Jalali 2017; Kiebert 1998 ‐ EORTC 22844; Klein 2002/Douw 2009; Prabhu 2014 ‐ RTOG 9802; Reijneveld 2016 ‐ EORTC 22033‐26033; Vigliani 1996); and two among people with grade 3 gliomas (Taphoorn 2007 ‐ EORTC 26951; Wang 2010 ‐ RTOG 9402). Of these nine studies, seven were randomised controlled trials (RCTs), which is to say that patients were randomly allocated to alternative treatments at recruitment. As the focus of this review is on long‐term outcomes, such outcome data derived from RCTs was from those subgroups of participants that survived and were able to complete long‐term assessments. Therefore, because participants with disease progression or who died were not followed up, the long‐term data from these trials are unlikely to be representative of the original randomised samples.
Two of the studies were observational (Klein 2002/Douw 2009; Vigliani 1996), with no attempt to randomly allocate participants to different treatments — patients receiving different treatments (physician or institution allocated) were simply followed up over time. Both of these studies reported outcomes in patients that had or had not received radiotherapy as part of their treatment for glioma.
Two studies (Klein 2002/Douw 2009; Taphoorn 2007) reported further long‐term data among survivors as a whole rather than by treatment group (Boele 2015; Habets 2014, respectively). We discuss these in more detail in the Agreements and disagreements with other studies or reviews section of the Discussion).
Numbers recruited and analysed
Altogether 2406 participants were recruited to the nine included studies. However, in all studies the number included in the analysis at various time points during follow‐up was generally considerably less than the total recruited or that had undergone baseline assessments. There was serious sample attrition due to death or disease progression, and there were missing data due to failure to carry out assessments or low participant response rates. In the Kiebert 1998 study the number followed up beyond two years was not clear; for the remaining studies, data were available for 503 participants (i.e. approximately a quarter of those recruited) at the final reported assessments, the timing of which varied between studies. In the Brown 2003 study, of 211 recruited there were follow‐up data for 97 at one year, 65 at two years and for only 38 at five years. Of 195 recruited to the Klein 2002/Douw 2009 study, data were available for 65 at the study end point with follow‐up times varying considerably for individual participants. Of 254 randomised in Prabhu 2014, MMSE data were available for 131 participants at one year and for 126 at two years; while for the 477 participants in Reijneveld 2016, data were reported for 253 at one year, 172 at two years and for 117 at three years. In Taphoorn 2007, of 268 randomised 149 were alive and progression‐free at 2.5 years and data were available for 94 of these patients. At four years there were data for 11 out of 31 participants in the Vigliani 1996 study; and at five years, of the 291 randomised in the Wang 2010 trial only 29 neurocognitive assessments were available. Finally, we included one study that recruited children, adolescents and young adults (Jalali 2017). This study included some participants with other types of brain tumour including craniopharyngioma, although the majority had glioma. We were only able to include data for a relatively small proportion of the sample; while 200 were recruited, only 66 were aged over 16 years and are included in the review, and at five years (the time point reported) only 23 provided outcome data.
Location of studies
Three studies were international and recruited patients from institutions in several countries (Kiebert 1998; Reijneveld 2016; Taphoorn 2007). The study by Wang 2010 was conducted in hospitals in the USA and Canada. The trials by Brown 2003 and Prabhu 2014 were carried out in the USA, and the remaining studies were carried out in the Netherlands (Klein 2002/Douw 2009), India (Jalali 2017), and France (Vigliani 1996).
Dates of recruitment
Participants were recruited to the various studies between 1985 and 2012, and follow‐up in some of the later studies continues. Three studies began recruitment in the 1980s (Brown 2003 1986 to 1996; Kiebert 1998 1985 to 1991; Vigliani 1996 1989 to 1993); four in the 1990s (Klein 2002/Douw 2009 1997 to 2000; Prabhu 2014 1998 to 2002; Taphoorn 2007 1996 to 2002; and Wang 2010 1994 to 2002); and two studies started recruitment after 2000 (Jalali 2017 2001 to 2012; Reijneveld 2016 2005 to 2012). In some studies recruitment was over a long period and it is possible that screening and diagnosis techniques, research personnel, aspects of care and adjuvant therapies changed over the course of the study.
Funding and conflict of interest
In the Kiebert 1998 study sources of funding were not reported. For the rest, all studies reported being financially supported by government, cancer charities or higher education research grants (Brown 2003; Jalali 2017; Klein 2002/Douw 2009; Prabhu 2014; Reijneveld 2016; Taphoorn 2007; Vigliani 1996; Wang 2010). In addition, three trials reported that they had received some support from commercial or private institutes (Klein 2002/Douw 2009; Reijneveld 2016; Taphoorn 2007). While in the Prabhu 2014 study it was reported that there was no commercial funding, several of the investigators reported receiving compensation from commercial organisations, although it was not clear whether this funding related to the reported work. Where conflict of interest was mentioned, no study authors reported conflict of interest other than Prabhu 2014 as stated above.
Characteristics of study participants
Age
All but one of the studies recruited only adult participants (> 18 years, although Kiebert 1998 recruited adults > 16 years). One study recruited children, adolescents and young adults up to the age of 25 years (Jalali 2017); approximately a third of the sample in this study were over 16 years and we have only included these young adults in our data and analysis. Vigliani 1996 had an upper age limit of 60 years for participants, whereas the remaining studies included older adults. For the eight studies recruiting adults, the median age of participants was between 40 and 49 years.
Gender
In most studies there was a larger proportion of male to female participants (approximately 60:40); In Kiebert 1998 and Vigliani 1996 there were similar numbers of men and women recruited.
Type and grade of glioma
Most studies recruited patients with low‐grade glioma and had criteria that excluded patients with other serious disease (e.g. other cancers or serious heart, liver or renal problems).
Five studies recruited participants affected by grades 1 or 2 supratentorial glioma including astrocytoma, oligodendroglioma or mixed disease (Brown 2003; Kiebert 1998; Klein 2002/Douw 2009; Prabhu 2014; Reijneveld 2016). Vigliani 1996 reported recruiting patients with grade 2 or 3 glioma (and in this non‐randomised study there was disparity between treatment groups in the type and grade of disease); and in the Wang 2010 and Taphoorn 2007 trials, participants had grade 3 disease and this was reflected in the poorer prognosis for patients in these studies compared with others. Finally, in the study recruiting children and young adults the sample included low‐grade glioma but also other types of brain tumours (Jalali 2017).
Surgical interventions
In all studies, most of the included patients had undergone surgical intervention prior to radio or chemo‐therapy although the proportions undergoing biopsy, partial or total resection varied. In the Jalali 2017 study the number of participants having surgery, and the type of surgical intervention, was not clear. In the remaining studies the proportions in treatment groups undergoing the different interventions was similar, except for the non‐randomised studies by Klein 2002/Douw 2009 and Vigliani 1996. In these observational studies, there was disparity between treatment groups in the numbers undergoing different surgical interventions and in the light of these differences in patient characteristics, between‐groups findings should be interpreted with particular caution. We have provided more information of the numbers undergoing surgery in the Characteristics of included studies tables.
Treatment with anti‐epileptic drugs
Only one of the included studies reported on the number of participants receiving anti‐epileptic drugs. In the Klein 2002/Douw 2009 study, 71% of patients in each of the treatment and control groups, respectively, received medication to prevent seizures.
Comparisons
The nine included studies examined a range of five different comparisons, as follows.
-
Radiotherapy versus no adjuvant treatment
-
Radiotherapy versus chemotherapy
-
High‐dose versus low‐dose radiotherapy
-
Standard versus stereotactic conformal radiotherapy
-
Chemoradiotherapy versus radiotherapy alone
1. Radiotherapy versus no adjuvant treatment
Two studies are included in this comparison and both used observational study designs (Klein 2002/Douw 2009; Vigliani 1996). In Vigliani 1996 allocation was by physician choice and in the retrospective study by Klein 2002/Douw 2009 there was no information on how allocation was made. The latter study involved a simple comparison between those participants that had or had not been treated with radiotherapy during the study period.
In the Klein 2002/Douw 2009 study the total mean radiotherapy dose was 55.6 Gy (standard deviation (SD) 6.1) with a fractional dose of 1.8 Gy to 2 Gy in 86 of the 104 participants. However, in 18 participants the fractional dose was greater than 2 Gy. The control group were patients with glioma who did not receive radiotherapy.
In the Vigliani 1996 study the radiotherapy dose was 54 Gy to 55.8 Gy in 1.8 Gy fractions over 6 weeks.
2. Radiotherapy versus chemotherapy
Reijneveld 2016 examined outcomes in participants randomised to either receiving radiotherapy (50.4 Gy in 28 fractions of 1.8 Gy up to 6.5 weeks) versus oral temozolomide daily for 21 out of 28 days repeated for up to 12 cycles (hence the duration of treatment was quite different in the two experimental groups).
3. High‐dose versus low‐dose radiotherapy
Two randomised studies examined higher versus lower total doses of radiotherapy; in both studies, although the fractional doses in the two arms were the same, the treatment period was longer in the higher dose groups (Brown 2003; Kiebert 1998). In Brown 2003 the total dose in the higher dose group was 64.8 Gy in 36 fractions over seven weeks compared with 50.4 Gy in 28 fractions over five and a half weeks weeks in the lower dose group. In Kiebert 1998 the higher dose was 59.4 Gy over six weeks compared with a lower dose of 45 Gy over five weeks.
4. Standard versus stereotactic conformal radiotherapy
This comparison included only one study that mainly recruited children under 16 years of age, but included a subgroup of participants between 16 and 25 years old (Jalali 2017). The radiotherapy dose in both arms was 54 Gy in 30 fractions over six weeks.
5. Chemoradiotherapy versus radiotherapy alone
Three studies examined the effects of chemoradiotherapy versus radiotherapy alone. In all studies the chemotherapy regimen comprised procarbazine, lomustine and vincristine. Radiotherapy was the same in both arms of each trial, although the dose used in the two studies was different. In the Prabhu 2014 trial, the radiotherapy dose was a total of 54 Gy in 30 fractions of 1.8 Gy over six weeks; while in the Taphoorn 2007 and Wang 2010 studies, the total dose was 59.4 Gy in fractions of 1.8 Gy. As in the Reijneveld 2016 study above, the duration of chemotherapy meant that the treatment period was more protracted in the chemoradiotherapy arms.
Outcomes and follow‐up
In this review, we aimed to include studies that reported longer term (two years or longer) neurocognitive or quality of life outcomes (or both). Several of the studies reported cognitive changes or impairment using the MMSE (Mini Mental State Examination) (Brown 2003; Prabhu 2014; Reijneveld 2016; Wang 2010). An MMSE score of 26 or lower out of 30 was the threshold applied as indicative of neurocognitive impairment in most of these studies. For the rest, Jalali 2017 collected data on intelligence quotients (for participants < 16 years), memory (Wechsler Memory Scale for participants > 16 years), and anxiety and depression; Vigliani 1996 used a battery of 12 neuropsychological tests and patients were considered globally deteriorated or improved when at least eight of 12 items were significantly modified by more than one standard deviation; and Klein 2002/Douw 2009 reported cognitive disability defined as deficits in at least five of 18 applied neuropsychological tests. Kiebert 1998, Reijneveld 2016, Taphoorn 2007 and Wang 2010 reported quality of life (QoL) outcomes.
Periods of follow‐up varied in these studies; while all studies followed up participants beyond two years, the number of participants at each progressive follow‐up point was reduced due to death or disease progression. For example, Brown 2003 followed up participants for a mean of 7.4 years but by this time more than half of the original sample had died (101/203 alive). In the Wang 2010 trial, median follow‐up time was 6.9 years in the surviving participants but 64% of participants had died, and in Taphoorn 2007 data from 2.5 years after radiotherapy were reported, by which point 59% of the original sample had died, and data on 32 of the long‐term survivors were reported in 2014. In Klein 2002/Douw 2009 the median follow‐up period was 12 years but the treatment groups were assessed at different time points and, in the intervention group, (radiotherapy) participants had received radiotherapy up to 20 years previously making results difficult to interpret. Vigliani 1996 reported outcomes up to four years after treatment. Kiebert 1998 followed up participants annually from two years; however, the published report contained QoL outcome data for participants between seven and 15 months only, which could not be used for our review purposes.
Excluded studies
After initial screening and full assessment of study reports we excluded 43 studies from the review. Fifteen studies were excluded as they did not assess or report neurocognitive or quality of life outcomes (Buglione 2014; Cairncross 2006; Combs 2008; Dai 2011; Ding 2017; Ediebah 2015; Eyre 1993; Goda 2017; Karim 2002; Malmstrom 2017; MRC 2001; Satoer 2014; Thomas 2001; van den Bent 2006; Wick 2009). Frequently in these studies the outcomes of interest were survival and disease progression. In eight studies all participants or a large proportion had high‐grade glioma such as glioblastoma, and in those studies where some participants had lower grade glioma separate results were not reported for these patients (Ali 2018; Chung 2018; NCT02655601; Repka 2018; Sichez 1996; Wheeler 2016; Wirsching 2018; Zhu 2017). In the study by Williamson 2017, participants had recurrent glioma and were undergoing re‐irradiation after initial treatment; this study also included participants with glioblastoma. Packer 2002 looked at a paediatric population which is outside the remit of this review.
Other important reasons for exclusion related to study design or the way results were reported. There were five observational studies with no comparator arm (Anand 2012; Armstrong 2002; Gregor 1996; Shaw 2006; Taylor 1998); and in a further five studies the control groups were not relevant to the aims of the review (e.g. the comparator group were healthy controls or had other types of disease or malignancy (Archibald 1994; Corn 2009; Costello 2004; Johannesen 2003; Sherman 2016)). In the study by Correa 2008 that included participants that had received radiotherapy, results were not analysed or reported separately for the radiotherapy arm which made results difficult to interpret. Issues relating to study design and sample selection also meant that results in Surma‐aho 2001 were difficult to interpret and likely to be at high risk of bias.
Finally six studies were excluded as they did not report original study data but were either reviews, commentary or letters to journal editors (Behrend 2014; Brown 2003b; Brown 2009; Klein 2004; Lunsford 2001; Taphoorn 1994); these reports may have included reference to studies already included or excluded from the review.
Risk of bias in included studies
See Figure 3 for the risk of bias summary table.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Allocation
In the seven RCTs, the methods used to randomise participants to experimental groups were mainly low risk or not clearly described. In four studies randomisation was carried out centrally and in these studies we assessed sequence generation and allocation concealment as low risk of bias (Jalali 2017; Kiebert 1998; Reijneveld 2016; Taphoorn 2007). In Brown 2003, there was probably centralised randomisation but this was not entirely clear. In the studies by Prabhu 2014 and Wang 2010, methods of sequence generation and allocation concealment were not well described (assessed as unclear risk of bias for both domains).
In the two non‐randomised studies there was likely to have been high risk of selection bias. In Vigliani 1996 allocation was down to physician choice and there were disparities between groups in terms of patient characteristics. For Klein 2002/Douw 2009, again without random allocation there was a likelihood of bias although methods were not described.
Blinding
In the randomised trials, blinding staff and participants was generally not feasible as treatment regimens in groups were different. Blinding of outcome assessment was also likely to have been at high risk of bias due to lack of blinding, as in this review we focus on subjective outcomes. Four studies were assessed as high risk of bias for both performance and detection bias (Prabhu 2014; Reijneveld 2016; Taphoorn 2007; Wang 2010). Brown 2003 and Kiebert 1998 did not mention blinding and it was unclear whether there was any attempt to blind those collecting outcomes to treatment allocation. Jalali 2017 had no treatment masking but reported that investigators collecting outcome data were unaware of treatment group.
In the non‐randomised studies there was no blinding (Klein 2002/Douw 2009; Vigliani 1996). In Klein 2002/Douw 2009, participants may have been unaware that their data were being used in a study as data were collected as part of clinical assessment. In Vigliani 1996, physicians chose treatment and recorded outcomes.
Incomplete outcome data
All of these studies were assessed as being at high risk of bias for sample attrition (defined by attrition of more than 20%; see Appendix 2). In Wang 2010 there was serious sample attrition but the investigators attempted to take sample loss into account in their analysis. For the rest, by two to three years following treatment there was a significant loss to follow‐up (with a half or more of the sample suffering disease progression or death). In Prabhu 2014 and Reijneveld 2016 there was considerable sample loss and at some assessment points there were different response rates in the two arms of these trials.
Selective reporting
Selective reporting bias is not easy to assess and this is reflected in our judgements, with all of the randomised studies being assessed as unclear risk of bias for this domain.
For the non‐randomised studies, we assessed Klein 2002/Douw 2009 as unclear risk of bias and Vigliani 1996 as high risk of bias.
Other potential sources of bias
Risk of other bias was generally not clear. In the Taphoorn 2007 trial progression‐free survival was better in one of the treatment arms and this may have affected some outcomes. In the non‐randomised studies there were baseline differences between groups (Klein 2002/Douw 2009; Vigliani 1996).
Effects of interventions
See: Summary of findings for the main comparison
We were not able to combine results in meta‐analysis due to differences in treatment comparisons, time points of follow‐up, and the different outcomes reported. However, we have entered data applicable to primary outcomes on forest plots, without totals, for narrative synthesis purposes. Due to the paucity of data available we have produced a single 'Summary of findings' table covering several different comparisons (summary of findings Table for the main comparison). The table includes dichotomous data for our primary outcome (neurocognitive impairment) for all but one of our comparisons (the Jalali 2017 study mainly recruited children and the limited data we summarise in the text below is for a subgroup aged over 16 years). We did not include estimates of absolute risk as part of our 'Summary of findings' table; this was because we considered that such estimates could be misleading. Findings reported in the review were based on subsets (progression‐free survivors) of samples originally recruited. As sample sizes at follow‐up tended to be small and event rates for outcomes low, there was considerable uncertainly in effect estimates. Absolute risks would reflect these serious uncertainties in the relative effect of interventions and were unlikely to be helpful in the interpretation of findings.
We had intended producing a 'Summary of findings' table for outcomes relating to quality of life but there were insufficient data to create a meaningful summary.
Primary outcomes
Cognitive impairment at 2 years or more after diagnosis/treatment
A. Radiotherapy versus no radiotherapy
A.1. Radiotherapy versus no adjuvant treatment
Two observational studies contributed data (Klein 2002/Douw 2009; Vigliani 1996), with the Klein 2002/Douw 2009 study authors reporting two time points up to 12 years after diagnosis/treatment, and Vigliani 1996 reporting the results of a battery of cognitive functioning tests from follow‐up up to four years after diagnosis/treatment. In the Klein 2002/Douw 2009 cohort, at the 12‐year follow‐up, the risk of cognitive impairment (defined as cognitive disability deficits in at least five of 18 neuropsychological tests) was greater in the radiotherapy group; at five to six years the difference between groups did not reach statistical significance (at five to six years, RR 1.38, 95% CI 0.92 to 2.06; n = 195; at 12 years, RR 1.95, 95% CI 1.02 to 3.71; n = 65) (Figure 4). In the Vigliani 1996 study, one study subject in the radiotherapy group had cognitive impairment at two years compared with none in the control group. We judged the evidence from these observational studies suggesting a possible negative relative effect of radiotherapy on long‐term cognitive impairment to be of very low certainty.
Forest plot of comparison A. Radiotherapy versus no radiotherapy, outcome: Neurocognitive impairment at 2 or more years after treatment. (dichotomous data)
With regard to neurocognitive scores, in the later Klein 2002/Douw 2009 study report (Douw 2009), the radiotherapy group had significantly worse mean executive functioning, attentional functioning and processing speed than the group that received no radiotherapy (Figure 5) and psychomotor functioning, verbal memory and working memory were reported as not significantly different; however, the raw data of the non‐significant findings were not given. In Vigliani 1996, there were no clear differences in any of the various cognitive outcomes measured at two years (n = 31) and four years (n = 15) after diagnosis. We judged this evidence as very low certainty due to inconsistency between these studies.
comparison A. Radiotherapy versus no radiotherapy, outcome: Neurocognitive impairment at 2 or more years after treatment. (continuous data)
A.2. Radiotherapy versus chemotherapy
One RCT contributed data on cognitive impairment, assessed at three years after randomisation (Reijneveld 2016). There was no clear difference in the proportion of participants with cognitive impairment between the trial arms at this time point (RR 1.43, 95% CI 0.36 to 5.70, n = 117) (Figure 4). MMSE scores were also measured at different time points and changes from baseline in MMSE scores up to 36 months were presented in a graph, with authors reporting that "no significant difference was recorded between the groups for the change in MMSE scores during the 36 month follow up” (p1533). Sparse data due to attrition and the wide 95% CIs for findings (imprecision) led us to judge the certainty of this evidence as low.
B. High‐dose radiotherapy versus low‐dose radiotherapy
Only one of the two studies — Brown 2003 and Kiebert 1998 — reporting this comparison contributed data. In the observational Brown 2003 study, only a small proportion of study subjects experienced a clinically significant decrease in MMSE score from baseline (more than 3 points) at the 2‐ and 5‐year follow‐ups and there were no clear differences between high‐ and low‐dose radiotherapy arms at either time point (Figure 6). Only 38 subjects of the original cohort of 203 contributed data at the 5‐year follow‐up. We judged the certainty of this evidence of no difference to be very low.
Forest plot of comparison B: High dose versus low dose radiotherapy, outcome: 2.1 Neurocognitive impairment at 2 years or more after treatment.
C. Conventional radiotherapy versus stereotactic conformal radiotherapy
One study involving younger people with low‐grade glioma contributed limited data from the subgroup aged 16 to 25 years (Jalali 2017). The numbers of participants with neurocognitive impairment at five years after treatment, assessed by the Wechsler Memory Scale, were two out of 12 versus none out of 11 participants in the conventional radiotherapy and conformal radiotherapy arms, respectively (Figure 7). These findings are inconclusive because the study was not powered to detect a difference in this subgroup of its participants (RR 4.62, 95% CI 0.25 to 86.72; n = 23; low‐certainty evidence).
Forest plot of comparison C: Conventional versus stereotactic conformal radiotherapy, outcome: Neurocognitive impairment at 2 years or more after treatment.
D. Chemoradiotherapy versus radiotherapy
Two RCTs reported cognitive impairment based on MMSE measurements for this comparison.
Prabhu 2014 defined it as a decline (of more than 3 points in MMSE score) in cognitive state compared with baseline and reported comparative data for 2‐year (110 participants), 3‐year (91 participants), and 5‐year (57 participants) time points (Figure 8); these dichotomous data showed no clear difference between the two study arms at any time point.
Forest plot of comparison D: Chemoradiotherapy versus radiotherapy, outcome: Neurocognitive impairment at 2 years or more after treatment.
Wang 2010 reported little raw data. Graphic representation of mean MMSE scores up to five years suggested that there was little difference between groups at any time point and authors reported that there was no difference between MMSE scores between the two study arms (P = 0.4752). Those assessed at two, three, four, and five years numbered 126, 110, 69 and 53 survivors respectively in this study. Only 29 out of 191 had completed all assessments at five years for the assessment of cognitive function (MMSE). Authors also reported that the group that received chemoradiotherapy had improving MMSE scores after two years, whereas in the radiotherapy‐only group mean MMSE scores among survivors remained constant over time.
We judged the findings (of no difference) to be low‐certainty evidence due to risk of (attrition) bias and imprecision.
A 2014 Taphoorn 2007 publication reported no difference in cognitive impairments between surviving patients treated initially with radiotherapy and chemoradiotherapy (7 and 20 patients, respectively) at a median survival of 147 months; however data were not reported separately.
Quality of life
We found no data on this outcome for comparisons A to C.
D. Chemoradiotherapy versus radiotherapy
Two RCTs involving people with grade 3 gliomas reported quality of life outcomes for this comparison. One reported no differences in Brain‐QoL scores between its two study arms over a 5‐year follow‐up period ( P = 0.2767; no raw data were given and denominators are not stated) (Wang 2010).
The other trial reported that the long‐term results of health‐related quality of life (HRQoL) "showed no difference between the arms" but did not give the raw data for overall HRQoL scores (Taphoorn 2007). However, authors reported appetite loss, fatigue, nausea and vomiting, physical functioning and drowsiness QoL mean scores; and at 2.5 years after radiotherapy there was no difference between the groups for any of these HRQoL components among participants with data at this time point (55 in the chemoradiotherapy arm and 39 in the radiotherapy arm).
We graded this evidence as low certainty because participants with progressive disease were excluded from assessment in these studies. Survivors in the chemoradiotherapy arms outnumbered the survivors in the radiotherapy arms and those with disease progression would be expected to experience a worse HRQoL; therefore the findings are biased towards no difference when there might be one.
Secondary outcomes
None of the review's secondary outcomes were reported.
Discussion
Summary of main results
We included nine studies altogether: these compared radiotherapy versus no adjuvant treatment (2 observational studies), radiotherapy versus chemotherapy (1 RCT), high‐dose radiotherapy versus low‐dose radiotherapy (subgroup analysis of patients without disease progression from 2 RCTs), conventional radiotherapy versus stereotactic conformal radiotherapy (1 RCT) and chemoradiotherapy versus radiotherapy (3 RCTs). All studies except for those of chemoradiotherapy versus radiotherapy involved people with low‐grade gliomas; whereas two of the chemoradiotherapy trials involved people with grade 3 glioma. As review outcomes are long‐term outcomes (2 or more years after treatment), attrition was high in most studies and, even in the RCTs, long‐term data were observational because the benefits of randomisation were lost through attrition. We did not perform meta‐analysis because the studies reported different time points and outcomes; however, where possible we entered data into forest plots to facilitate narrative synthesis, evidence grading and discussion.
Radiotherapy versus no radiotherapy
For cognitive impairment at two or more years after treatment (measured as a categorical variable in 3 studies), limited evidence suggested that radiotherapy may increase the risk of long‐term cognitive impairment; however the magnitude of this effect was not estimable and we graded the evidence as 'very low certainty'. Evidence on the associated continuous variables that comprised different components of cognitive functioning were also very low certainty. We found no comparative data on quality of life.
High‐dose versus low‐dose radiotherapy
Only one study contributed data on cognitive impairment at two and five years after treatment and its findings showed no difference between these radiotherapy options (very low certainty evidence).
Conventional radiotherapy versus stereotactic conformal radiotherapy
Low‐certainty evidence suggested there may be little or no difference in cognitive impairment at five years after randomisation between these options.
Chemoradiotherapy versus radiotherapy alone
Low‐certainty evidence suggested there may be little or no difference between these options in cognitive impairment among survivors at two and five years' follow‐up. The evidence also suggested that there may also be little or no difference in quality of life at two years or more among glioma survivors who receive either treatment option (low‐certainty evidence).
We identified no relevant data on the review's secondary outcomes or for the brief economic commentary.
Overall completeness and applicability of evidence
With regard to radiotherapy versus no radiotherapy, the included studies were fairly old so this very low to low‐certainty evidence of an increased risk of cognitive impairment might not be applicable to modern radiotherapy techniques, such as image‐guided and conformal radiotherapy, which aim to reduce radiation exposure to normal tissue.
Findings on the cognitive effects of high‐dose versus low‐dose radiotherapy were inconclusive; however, in these studies, death rates and toxicity rates were slightly but consistently higher in the high‐dose arms. As high‐dose radiotherapy in low‐grade glioma is not advocated, further studies on this are unlikely.
Evidence from studies of radiotherapy versus chemotherapy or chemoradiotherapy suffered from high attrition and insensitive measurement tests. Data from various studies employing better measurement tests are not yet mature (Klein 2017).
With regard to conventional radiotherapy versus stereotactic conformal radiotherapy, we derived evidence relating to the review's primary outcomes from a group of young participants (aged 16 to 25 years) in only one trial and, unfortunately, the findings were underpowered to be conclusive. Further research among adult populations with low‐grade glioma would be of interest. Whilst neuroendocrine dysfunction was measured in this study, we were unsuccessful in obtaining separate data from the authors for the subgroup of patients older than 16 years with glioma.
We were unable to synthesise evidence on secondary review outcomes due to a lack of data.
Quality of the evidence
The main review results suggesting that radiotherapy may have a negative effect on cognitive functioning in the long term should be interpreted with caution because the quality of the evidence is low.
Evidence on cognitive function was most commonly derived from study data collected using the MMSE, which lacks sensitivity to mild changes in cognitive impairment and changes due to focal lesions. This is an important limitation of the evidence, as neurocognitive problems related to brain tumours can be subtle or restricted to certain neurocognitive domains only (as suggested by the Klein 2002/Douw 2009 data), depending on their location (Day 2016).
Brief economic commentary
To supplement the main systematic review of the long‐term complications of radiotherapy in those with glioma we sought to identify economic evaluations which included the long‐term effects of radiotherapy as part of the evaluation. No economic studies were identified that analysed the long‐term consequences of radiotherapy. The apparent shortage of relevant economic evaluations indicates that economic evidence regarding the long‐term effects of radiotherapy on long‐term glioma survivors is needed.
Potential biases in the review process
Whilst we did not pool data, it might have been reasonable to do so for the primary dichotomous outcome 'Cognitive impairment at 2 years or more after treatment' of the 'Radiotherapy versus no radiotherapy' comparison. We included three studies in this forest plot, two comparing radiotherapy with no adjuvant treatment and one comparing radiotherapy with chemotherapy. We chose not to pool these data because of the clinical heterogeneity (different measurement time points and different control interventions). Had we done so (using the 5‐ to 6‐year data from Klein 2002/Douw 2009, not the 12‐year data; see Figure 9) the effect estimate in favour of no radiotherapy would have been an RR of 1.40 (95% CI 0.95 to 2.05). With downgrading for imprecision and risk of bias, we would most likely have graded this evidence as low certainty. Whilst our narrative synthesis does not provide an overall effect estimate, the grading and interpretation of the evidence is reasonably consistent with the latter.
Forest plot of comparison A (exploratory with totals): Radiotherapy versus no radiotherapy, outcome: Neurocognitive impairment at 2 or more years after treatment.
Other potential biases are as follows.
-
Neurocognitive impairment was variously measured across included studies. We extracted and analysed both dichotomous and continuous data where available. As it is possible for differences in mean scores between treatment groups to be statistically significant but not clinically meaningful, evidence on changes in continuous data (mean scores) should be — and were — interpreted with caution.
-
In the Klein 2002/Douw 2009 study, which reported 12‐year follow‐up data (Douw 2009), only the significant results were reported as raw data in the text. Psychomotor functioning, verbal memory, and working memory were not significant (data shown in graphs) but the results tended to be in the same direction (favouring the 'no radiotherapy' group). We did not attempt to obtain these numerical data from the authors as we considered that any data obtained would be of a very low quality and did not warrant the (investigator's) efforts required to retrieve it, given that the findings were at high risk of bias anyway.
-
We included Kiebert 1998, which compared high‐dose radiotherapy with low‐dose radiotherapy; however, the study ended up contributing no usable data to the review. Whilst the study methods stated that participants were followed up annually after 24 months, only data from participants between 7 and 15 months after diagnosis were reported in the published paper and we were unable to obtain any subsequent follow up data. Findings from the 7‐ to 15‐month assessment showed no significant difference in neurological impairment and no significant difference in the proportion of patients with the worst neurological scores (data were not shown in the paper). There was no significant difference in QoL scores overall but some QoL items were worse with high dose, namely emotional functioning and leisure time activities (P = 0.009 and P = 0.017, respectively). By not using these data, we might have missed an opportunity to estimate the effects of high‐dose versus low‐dose radiotherapy on quality of life outcomes.
-
Prabhu 2014, which compared chemoradiotherapy with radiotherapy, in addition to reporting neurocognitive decline, reported the numbers of participants in each group that experienced an improvement in cognitive functioning (based on a 3‐point increase in MMSE score) over a 5‐year period. Similarly, Vigliani 1996 reported cognitive improvement according to author‐specified criteria in two patients in this study following radiotherapy. As cognitive improvement was not a pre‐specified outcome, we did not present or analyse these data.
-
We included a trial of conventional radiotherapy versus stereotactic conformal radiotherapy, a more recent radiotherapy technique that aims to reduce the radiation exposure of normal tissue; however this trial was conducted mainly in younger people (Jalali 2017). The sample mainly comprised children and young people with glioma but also included other brain tumours including craniopharyngioma. Most results were not broken down by diagnosis or age group and we were unable to obtain additional data for the subgroup of interest from study authors. In this study, most participants were under 16 years but one‐third were aged 16 to 25 years, and we limited our data extraction to the older age group. Overall, however, neurocognitive (intelligence quotient or memory scores) of patients in the stereotactic conformal radiotherapy arm were either stable or showed an improvement over five years compared with patients in the conventional radiotherapy arm (difference in slope = 1.48; P = 0.04), which was the same direction of the neurocognitive effect reported for the older subgroup only, but for which the data were sparse. This trial also reported the incidence of new endocrine dysfunctions, which were significantly fewer in the stereotactic conformal arm compared with the conventional arm (52% versus 29%; P = 0.02); we did not use these data in the review, however, because they were derived mainly from patients under 16 years old and we were unsuccessful in obtaining subgroup data from the authors.
-
We excluded studies of glioblastoma because of the poor rates of survival at two years and more. However, with improved survival rates for IDH‐mutated gliomas, useful long‐term data might become available from such studies in the future and we might need to reconsider our study inclusion/exclusion criteria.
Agreements and disagreements with other studies or reviews
Two cohort studies (Boele 2015 and Habets 2014) reported additional long‐term data related to included studies (Klein 2002/Douw 2009 and Taphoorn 2007, respectively) for survivors as a group, rather than separately by treatment group. Both of these longer term studies compared HRQoL in glioma patients (low‐grade glioma (LGG) and grade 3 patients, respectively) with healthy controls and previous assessments. Habets 2014 also assessed cognitive functioning. In Boele 2015, the assessments were at around six and 12 years after diagnosis and initial treatment, respectively, and in Habets 2014 assessment was at a median of 147 months after diagnosis. Neither of these studies had comparative data that could be included in our meta‐analyses. Their main findings are as follows.
Boele 2015 found that LGG patients had lower physical role functioning (P = 0.004) and general health perceptions (P = 0.004), but no other statistically significant differences were observed compared with healthy controls. In the majority of patients both physical (87.7%) and mental (80%) HRQoL remained stable; however, the mean physical HRQoL score was reported to be significantly worse at 12 years than at six years (49.5 versus. 46.9, P < 0.01). Authors concluded that "although HRQoL remains mostly preserved in the majority of LGG patients, a subset of patients experience detectable decline on one or more HRQoL scales despite long‐term stable disease."
Habets 2014 reported findings of 32 out of 37 long‐term survivors of grade 3 gliomas who had participated in the Taphoorn 2007 trial comparing radiotherapy with chemoradiotherapy. The number of survivors was less than 10% of the original sample and about a third of those assessed at 2.5 years. Ten out of 32 patients evaluated had received radiotherapy initially, the rest had received chemoradiotherapy. Compared with healthy controls, survivors who had never progressed had lower working memory capacity, information processing speed, psychomotor functioning, attention, and executive functioning. Investigators reported that initial treatment did not correlate with HRQoL or cognitive functioning findings and that HRQoL in the long term for this cohort was similar to the findings at 2.5 years.
With such studies, it is important to bear in mind that differences between glioma patients and healthy controls could be related to the glioma itself, as well as to treatment. Also, long‐term data on cognitive effects and quality of life are inherently biased by the effect of a given treatment on survival. This is particularly relevant to studies that show substantial differences in survival, as quality of life would plausibly be worse among patients with a shorter survival time at a particular time point distant from treatment, and cognitive data would be influenced by the greater attrition of participants in the study arm with the shorter survival. Such a positive correlation between survival and quality of life is evident in the Wang 2010 study, which evaluated the effect of chemoradiotherapy versus radiotherapy alone among people with anaplastic oligodendrogliomas, and also in CATNON 2017 (see below).
Ongoing studies
The review process identified six ongoing studies that hopefully will contribute data to a future version of this review (Characteristics of ongoing studies). Ongoing randomised trials include long‐term neurocognitive outcomes of the EORTC study 22033‐26033 (radiotherapy versus temozolomide; Klein 2017); CATNON 2017 (radiotherapy versus radiotherapy plus adjuvant temozolomide and other comparisons ‒ 4 study arms); and NCT03180502 (proton beam versus intensity‐modulated radiotherapy). Ongoing observational studies include NCT00457210, NCT02544178 and NCT03055364; one of these was registered in 2007 and forthcoming data would seem unlikely at this stage. All the other studies are likely to report results from 2020 onwards.
Study flow diagram (date of search 16/02/18).
Study flow diagram (date of search 9/10/18).
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Forest plot of comparison A. Radiotherapy versus no radiotherapy, outcome: Neurocognitive impairment at 2 or more years after treatment. (dichotomous data)
comparison A. Radiotherapy versus no radiotherapy, outcome: Neurocognitive impairment at 2 or more years after treatment. (continuous data)
Forest plot of comparison B: High dose versus low dose radiotherapy, outcome: 2.1 Neurocognitive impairment at 2 years or more after treatment.
Forest plot of comparison C: Conventional versus stereotactic conformal radiotherapy, outcome: Neurocognitive impairment at 2 years or more after treatment.
Forest plot of comparison D: Chemoradiotherapy versus radiotherapy, outcome: Neurocognitive impairment at 2 years or more after treatment.
Forest plot of comparison A (exploratory with totals): Radiotherapy versus no radiotherapy, outcome: Neurocognitive impairment at 2 or more years after treatment.
| Long‐term neurocognitive and other side effects of radiotherapy, with or without chemotherapy, for glioma | ||||
| Patient or population: people with glioma surviving at least two years Settings: tertiary care | ||||
| Comparison and Outcomes | Relative effect | No of participants andstudies | Quality of the evidence | Comments |
| Intervention: radiotherapy Comparison: no adjuvant treatment Outcome: neurocognitive impairment at 5‐ to 6‐year follow‐up | RR 1.38 (0.92 to 2.06) | 1 study with data for 195 participants | ⊕⊝⊝⊝ | Outcome defined as cognitive disability deficits in at least 5 of 18 neuropsychological tests |
| Intervention: radiotherapy Comparison: no adjuvant treatment Outcome: neurocognitive impairment at 12 year follow‐up | RR 1.95 (1.02 to 3.71) | 1 study with data for 65 participants | ⊕⊝⊝⊝ | Outcome defined as cognitive disability deficits in at least 5 of 18 neuropsychological tests |
| Intervention: radiotherapy Comparison: no adjuvant treatment Outcome: neurocognitive impairment at 2 year follow‐up | RR 2.50 (0.11 to 56.98) | 1 study with data for 31 participants | ⊕⊝⊝⊝ | There was a single event for this outcome in this observational study. The outcome was defined as a significant deterioration (≥ 1 SD) in 8 out of 12 neuropsychological tests |
| Intervention: radiotherapy Comparison: chemotherapy Outcome: neurocognitive impairment at 3 year follow‐up | RR 1.43 (0.36 to 5.70) | 1 study with data for 117 participants | ⊕⊕⊝⊝ | Outcome defined as a MMSE score of 26 or less |
| Intervention: high‐dose radiotherapy Comparison: low‐dose radiotherapy Outcome: neurocognitive impairment at 2 years after treatment | RR 0.53 (0.06, 4.85) | 1 study with data for 65 participants | ⊕⊝⊝⊝ | Outcome defined as decrease in MMSE score from baseline (more than 3 points).There was serious and uneven attrition between groups in this study. |
| Intervention: high‐dose radiotherapy Comparison: low‐dose radiotherapy Outcome: neurocognitive impairment at 5 years after treatment | RR 0.16 (0.01 to3.20) | 1 study with data for 38 participants | ⊕⊝⊝⊝ | Outcome defined as decrease in MMSE score from baseline (more than 3 points). There was serious and uneven attrition between groups in this study. |
| Intervention: chemoradiotherapy Comparison: radiotherapy Outcome: neurocognitive impairment at 3 years after treatment | RR 0.37 (0.02 to 8.88) | 1 study with data for 91 participants | ⊕⊕⊝⊝ | Outcome defined as a decline (of more than 3 points in MMSE score) in cognitive state compared with baseline |
| Intervention: stereotactic conformal radiotherapy Comparison: radiotherapy Outcome: neurocognitive impairment at 5 years after treatment | RR 4.62 (95% CI 0.25 to 86.72) | 1 study with data for 23 participants | ⊕⊕⊝⊝ | Outcome defined as a decline (of more than 3 points in MMSE score) in cognitive state compared with baseline. There was serious sample attrition at 5 years. |
| GRADE Working Group grades of evidence | ||||
| Abbreviations: SD = standard deviation; MMSE = Mini Mental State Exam 1. Single study contributing data had very serious study design limitations (−2) | ||||
