Reporting of patient-reported health-related quality of life in adults with diffuse low-grade glioma: a systematic review

Abstract

Background

Patient-reported health-related quality of life (HRQoL) analysis can provide important information for managing the balance between treatment benefits and treatment-related adverse effects on quality of life (QoL). This systematic review sought to identify the range of HRQoL measures used for patients with diffuse hemispheric WHO grade II glioma (DLGG) and assess the quality of HRQoL reporting.

Methods

This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Databases were searched for full-text English articles reporting HRQoL outcomes in adult patients with DLGG.

Results

Eleven different QoL measures were used across the 26 included studies, none of which has been validated in patients with DLGG. Heterogeneity of study design prevented pooled analysis of data investigating the effect of interventions or establishing long-term HRQoL. Low rates of participation at baseline (mean: 64.0%) and high rates of subsequent dropout (2.1% per month) were identified. Five studies gave statistical methods to deal with missing data or provided evidence of clinical significance of HRQoL results.

Conclusions

The results demonstrate a paucity and heterogeneity of reporting of HRQoL in the DLGG literature, highlighting the need for a standardized assessment schedule and set of validated quality-of-life measures for future studies.

Diffuse low-grade glioma (DLGG), encompassing World Health Organization (WHO) grade II astrocytoma, oligodendroglioma, or oligoastrocytoma has an incidence of approximately 1:100 000 person years¹ and a median age at diagnosis ranging from 42 to 48 years of age.^2–4 While the vast majority of patients with DLGG will die from malignant transformation of the tumor, median survival from diagnosis is between 5 and 15 years.⁵ Thus, there is a significant period of time within which patients with DLGG can develop neurologic and systemic sequelae, either as a result of disease progression or from treatment that can impair quality of life (QoL).⁶ Further impairment in QoL can arise as a result of epilepsy, which is known to occur in 70%–90% of presentations of patients with DLGG.^7,8 Given the relatively young age at diagnosis, there may be a significant impact of these sequelae during the long overall survival of patients with a diagnosis of DLGG. As a result, considering QoL is recognized as being equally important as prolonging survival.^1,9

Patient-reported health-related quality of life (HRQoL) outcomes, while subjective, offer the most comprehensive representation of a patient's perceived QoL. HRQoL refers specifically to patients' perception of their physical and occupational function, psychological state, social interaction, and somatic sensation influenced by their medical condition and/or consequences of therapy.¹⁰ Such assessments can provide pivotal details in pursuit of the optimum “onco-functional balance” between prolonging survival and preserving QoL.^11,12 Furthermore, HRQoL outcomes may have prognostic significance⁵ or provide information on disease progression that can complement objective outcome measures such as overall and progression-free survival.¹³

HRQoL in the context of DLGG has not been systematically evaluated. Dirven et al. conducted a systematic review that assessed the level of reporting of HRQoL results in randomized controlled trials (RCTs) of patients with primary brain tumors; however, none of the studies in their review included patients with DLGG.⁹ A similar systematic review in 2002 identified 2 studies that included patients with low-grade glioma, of which only one utilized a HRQoL measure.^14–16 Although comprehensive reviews have been published on interventions in patients with DLGG and their effect on QoL,^5,17–19 no review has specifically assessed the reporting of HRQoL in patients with DLGG across randomized and nonrandomized studies with a systematic design.

This is important, given the relatively long overall survival of this population and the disparate temporal effects on HRQoL of potential therapies. For instance, surgical resection may result in acute worsening of HRQoL prior to a rapid improvement, consistent with temporary postoperative neurological deficits.²⁰ Conversely, there is evidence that patients with DLGG receiving radiotherapy demonstrate declining cognitive functioning over time.²¹ A systematic design enables a comprehensive overview of current practices with regard to the measurement and reporting of HRQoL; including nonrandomized studies is crucial, given the aforementioned absence of studies involving patients with DLGG with a randomized design.

This systematic review of all studies utilizing HRQoL measures in patients with DLGG had 4 specific objectives:

1. To establish and evaluate the HRQoL measures that have been used in studies of patients with DLGG.

2. To assess the level of heterogeneity of inclusion and exclusion criteria and time points of HRQoL assessment that would facilitate or limit pooled analyses.

3. To measure the levels of participation, dropout and missing items, and strategies to reduce these in studies with longitudinal measurement.

4. To evaluate the level of HRQoL reporting in accepted studies relative to guidelines published by the International Society for Quality of Life Research (ISOQOL) on the minimum reporting standards.²²

This evaluation is designed to be the first step in the process of developing a standardized schedule and set of HRQoL measures for DLGG. This will be developed via a methodologically robust consensus process (eg, a Delphi survey) involving all major stakeholders in DLGG including neurosurgeons, neuro-oncologists, academics, QoL experts, patients, and carers.

Methods

This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and has been registered with the PROSPERO international prospective register of systematic reviews (registration number CRD42014013363).

Search Strategy

The following databases were independently searched on March 10, 2015 by 2 authors (DF and AC): MEDLINE via Ovid, EMBASE via Ovid, PsycINFO, the Cochrane Central Register of Controlled Trials (CENTRAL), and the Cochrane Database of Systematic Reviews (CDSR) via the Cochrane Library. Search terms are provided in the online supplementary material (Supplementary material, Table S1).

Initially, titles and abstracts were screened for relevant papers by 2 authors independently (DF and AC). Decisions were blinded; when disagreements occurred, both authors discussed the disparities and resolved them. The bibliographies of accepted papers were also examined for additional articles not identified in the systematic search. A full flow diagram for the search is shown in the online supplementary material (Supplementary material, Fig. S1). The inclusion criteria were as follows:

1. Study design: RCTs, controlled clinical trials, prospective, and retrospective observational studies were included conditional upon a sample size >10 patients. We excluded studies with smaller sample sizes on the grounds of a potential publication bias risk. Only full articles in English and published after 1990 were included, with articles in other languages, conference abstracts, and grey literature being excluded.

2. Population: All studies including adult (≥18 years) patients with confirmed diagnosis of hemispheric WHO grade II astrocytoma, oligodendroglioma, or oligoastrocytoma (DLGG), achieved through biopsy or postresection histopathology.

3. Intervention: All treatment modalities were included for comparative purposes.

4. Outcome: Patient-reported HRQoL assessment

Data Collection

Data extraction was performed independently by 2 authors (DF and DA). Forward and backward searches of included studies provided 2 further articles not identified in the database search. When inconsistent reporting formats of data were published, authors were contacted with a standardized table to collect all data across accepted studies.

HRQoL Measures, Assessment Schedule, Inclusion, and Exclusion Criteria

To assess the level of heterogeneity of HRQoL reporting, the measures used, time points for data collection, and study inclusion/exclusion criteria were collected and compared. For all HRQoL measures used in at least 2 studies, mean and standard deviation HRQoL data were collected to facilitate assessment of whether or not pooled analyses could be conducted using the current literature data. In order to collect specific HRQoL data for patients with DLGG, only studies publishing data specifically for this population (or providing this data following correspondence with the authors) were included.

Missing Data

In total, 3 levels of missing data were evaluated: rates of participation in studies where recruitment of patients for HRQoL assessment was described, level of dropout across all time points in longitudinal studies, and level of missing items in completed questionnaires. Sample sizes across all time points were collected in order to calculate longitudinal dropout rates and calculated as a percentage per month of follow-up to control for different follow-up time periods. Specific reasons for lack of participation and dropout were collected when possible. Furthermore, data related to the reasons for missing items in submitted forms were collected, as were associated statistical methods related to all levels of missing data.

ISOQOL Standards

In order to improve implementation and reporting of HRQoL measures, the International Society for Quality of Life Research (ISOQOL) published a set of minimum standards for patient-reported QoL outcomes in RCTs in 2013. These standards were developed through an established consensus process involving a number of different stakeholders including patients.²² Dirven et al. applied the ISOQOL guidelines in the context of brain tumors, splitting one item and including 3 additional items for this patient population.⁹ Studies were evaluated against these 22 items, with those evaluating HRQoL as a primary outcome assessed against an additional 11 items as recommended in the original set of minimum standards,²² thus providing a total of 33 items for analysis. Evaluation was performed independently by 2 authors (DF and DA). When disagreements occurred, both authors discussed the disparities and resolved them. Items specific to RCTs were excluded for assessment of nonrandomized study designs.

Results

Demographics

The 26 accepted studies provided a total sample of 2,636 patients. All studies were included in evaluation of HRQoL measures used, reporting schedules, inclusion and exclusion criteria, and ISOQOL standards. Data relating to participation, dropout, and missing items were available for 19 of 26 studies. HRQoL data were collected for 13 of 26 studies to assess the feasibility of pooled analysis.

Study Design

Three studies were RCTs. There were 5 prospective trials, 3 case-control studies, 2 cohort studies, 5 case series, and 8 cross-sectional studies. Interventions included surgery alone, temozolomide chemotherapy, radiotherapy, combination therapy, cognitive rehabilitation, and acupuncture. Comparators included healthy controls, patients with non-Hodgkin's lymphoma and chronic lymphatic leukemia, and the patients′ next of kin; further comparisons within cohorts were based on tumor size, location, and timing of surgery (early resection vs biopsy and wait). Details of each study are provided in the online supplementary material (Supplementary material, Table S2).

Quality of Life Measures

A total of 11 different metrics were used to establish QOL, not including performance measures such as Karnofsky Performance Status (KPS). Details of the number of studies per measure and further details of measures used in more than one study are shown in Table 1. None of the measures were validated specifically in the DLGG population. The EORTC QLQ-BN20, used in 42.3% of studies, was validated in a population of patients with anaplastic oligodendroglioma and glioblastoma,²³ while the FACT-Br, used in 11.5% of studies, was validated in a cohort of patients with brain tumors in which 7 patients had a mixed glioma.²⁴ All measures were validated in the language in which the instrument was used. The EORTC QLQ-BN20 was the only HRQoL measure to specifically include domains related to epilepsy, while the EORTC QLQ-C30 was the only HRQoL measure that evaluated cognitive functioning as a specific domain. Twelve studies reported the KPS for their cohorts; 4 studies noted that KPS is an inadequate measure of HRQoL compared with patient-reported measures.^25–28

Table 1.

Health-related quality of life measures identified in the systematic review

QoL Measure	Studies/n	DLGG/n	Domains	Target	Scale (MCID)	Reliability/validity
EORTC QLQ-BN20	11	499	Future uncertainty, visual disorder, motor dysfunction, communication disorder Symptoms (headache, seizure, drowsiness, hair loss, itchy skin, lower limb weakness, bladder control)	Patients with brain cancer	0–100 (5–10)44	Evidence of validity in 891 brain tumor patients for forms in 12 languages.23 Cronbach's α = 0.70–0.87 Test-retest ICC = 0.23–0.81 (9/12 >0.50)51
EORTC QLQ-C30	8	255	Global QoL scale Functioning (physical, role, cognitive, emotional, social) Symptoms (fatigue, pain, nausea/vomiting, dyspnea, appetite loss, sleep disturbance, constipation, diarrhea) Financial impact	Patients with cancer	0–100 (5–10)44,52	Validated in cohort of 305 patients across 3 language-cultural groups with lung cancer. Cronbach's α = 0.54–0.89 (during treatment 8/9 >0.70)53
SF-36	7	394	Physical activities, social activities, role limitations due to physical health, bodily pain, mental health, role limitations due to mental health, vitality (energy and fatigue), general health problems	Generic questions used to compare burden of diseases and benefits of treatments, not specific to any particular disease or patient group.54	0–100 (N/A)	Validated in cohort of 3,445 patients across 24 subgroups in English and in French.55,56 Cronbach's α 0.65–0.94, median 0.85.55,57
FACT-Br	3	20	Physical well-being, social/family well-being, relationship with doctor, emotional well-being. functional well-being Brain subscale	Patients with brain tumors.	0–100 (N/A)	Validated in cohort of 101 patients with primary brain tumours.24
EuroQoL-5D	2	79	Mobility, self-care, usual activities, pain/discomfort, anxiety/depression	Range of conditions and treatments.58	-0.594–1 (N/A)	Validated in the Norwegian population.59
Sintonen 15D	2	0	Mobility, vision, hearing, breathing, sleeping, eating, speech, elimination, usual activities, mental function, discomfort and symptoms, depression, distress, vitality, sexual activity	Applicable to a wide range of conditions and treatments.60	0–1 (N/A)	Highly reliable, sensitive and responsive to change60
Other Measures	11	216	Affect Balance Scale (ABS) – 10-item instrument measuring positive affect or “happiness”61 Customized questionnaire – Used in Kiebert et al. as there was no established instrument to measure HRQoL at the time.15 FACT/NCCN FBrSI-15 – further questionnaires from the Functional Assessment of Cancer Therapy series (General and Fatigue-specific). Nottingham Health Profile – Instrument designed to determine perceived health status in the elderly.62 Subjective Estimation of Quality of Life (SQoL) – External life conditions, interpersonal relationships and internal psychological states.63

Abbreviation: MCID, Minimal clinically important difference.

Inclusion Criteria, Assessment Schedule, and Propensity for Pooled Analyses

Heterogeneity of inclusion criteria and HRQoL measures limited analysis of data in terms of assessing interventions (Table 2). Results for 3 studies evaluating the effect of temozolomide chemotherapy on HRQoL in patients with DLGG is provided in the online supplementary material (Supplementary Material, Fig. S2). Pooled analysis was not performed due to differing exclusion criteria and assessment schedules. Only one study evaluated the effect of radiotherapy alone. Comparisons between temozolomide and radiotherapy were not possible due to the lack of comparable longitudinal data specific for patients with DLGG.²⁹ Other interventions included cognitive rehabilitation,³⁰ image-guided surgery,²⁸ and acupuncture.³¹

Table 2.

Included studies with longitudinal health-related quality of life evaluation of intervention in patients with diffuse hemispheric WHO grade II glioma

Study	Inclusion/exclusion Criteria	Intervention	Treatment Status at Baseline	Years Since Diagnosis at Baseline/mean (SD)	HRQoL Measure
Chemotherapy
Blonski et al 201225	Inoperable supratentorial DLGG Age >18 y Previous radiotherapy patients excluded.	Temozolomide followed by resection	Biopsy: 7/10 One partial resection: 2/10 2 partial resections: 1/10 Radiotherapy: none	4.8 (1.4)	QLQ-C30 QLQ-BN20
Brada et al 200336	DLGG Age >18 y No previous treatment other than surgery Stable or progressive disease Suitable for temozolomide therapy Patients with sequelae secondary to raised intracranial pressure were excluded	Temozolomide chemotherapy	Biopsy: 18/30 One resection: 9/30 2 or more resections: 3/30 Radiotherapy: none Chemotherapy: none	N/A	QLQ-C30 QLQ-BN20
Pace et al 200335	DLGG MRI tumor progression KPS > 60 Bartel index >80	Temozolomide chemotherapy	Recurrence resection: 16/43 Radiotherapy: 30/43 PCV chemotherapy: 16/43	Median 5.1 (range 0.4–12.0)	QLQ-C30 QLQ-BN20
Radiotherapy
Shih et al 201564	DLGG MRI tumor recurrence/progression Tumor symptoms Age ≥18 y KPS ≥ 70 No cognitive deficits No prior cranial irradiation No comorbidities to suspect survival for <5 y	Proton therapy	Gross total resection: 4/20 Subtotal resection: 12/20 Biopsy only: 4/20	1.5 (0.1–13)	Beck Depression Inventory Beck Anxiety Inventory FACT-Br Employment status
Other
Gehring et al 200930	DLGG aged 18–70 y or anaplastic glioma aged <50 y No evidence of disease progression ≥6 mo before study KPS > 70 No treatment Severely cognitively impaired were excluded	Cognitive rehabilitation	Biopsy: 40/140 Resection: 92/140 Radiotherapy: 86/140 Chemotherapy: 15/140	Intervention: 5.2 (range 38.1) Control: 6.1 (range 28.3)	SF-36
Jakola et al 201128	Confirmed gliomas Age >16 y Severe aphasia/cognitive deficits excluded	Image-guided surgery	Surgery: 30/88	N/A	EuroQol-5D
Yu et al 201365	DLGG Hemiparesis Age 18–70 y ≥2 wk after brain surgery ≥4 wk after chemotherapy ≥6 wk after radiotherapy KPS 50–90 Life expectancy >3 mo	Acupuncture	Subtotal resection: 29/53 Gross total resection: 24/53	N/A	QLQ-C30 QLQ-BN20

Abbreviation: DLGG, diffuse low-grade glioma; HRQoL, health-related quality of life; PCV, procarbazine-CCNU-vincristine.

If N/A, information could not be ascertained from the manuscript or after requesting information from authors.

Heterogeneity of HRQoL measures and inclusion criteria was also observed in studies assessing long-term progression of HRQoL (Table 3). Results across function and symptom domains for the EORTC QLQ-BN20 and SF-36 HRQoL measures are provided in the online supplementary material (Supplementary Material, Figs S3 and S4); again, no pooled analysis was performed. There were wide variations in the proportions of patients who underwent tumor biopsies (range: 12.5%–46.1%, n = 5), tumor resections (range: 35.4%–74.7%, n = 5), and radiotherapy treatment (range: 49.2%–55.1%, n = 3). Heterogeneity was also observed in the period of follow-up, ranging from 4.4 ± 4.5 years to 12.8 ± 3.5 years after diagnosis.

Table 3.

Studies reporting pretreatment and posttreatment follow-up health-related quality of life in patients with diffuse low-grade glioma

Study	Inclusion/exclusion Criteria	Treatment Status	Years Since Diagnosis at Baseline/mean (SD)	HRQoL Measure
Posttreatment follow-up
Aaronson et al 201132	No clinical recurrence >1 y after diagnosis No radiologic signs within 3 mo of study Exclude corticosteroid use Exclude those with severe psychiatric/neurocognitive deficits	Biopsy: 71/195 Resection: 94/195 Radiotherapy: 103/195	5.6 (3.7)	SF-36 QLQ-BN20
Boele et al 201540a	As in Aaronson et al 2011	Biopsy: 19/65 Resection: 41/65 Radiotherapy: 32/65	12.8 (3.5)	SF-36 QLQ-BN20
Jakola et al 201411	Supratentorial DLGG	Biopsy: 20/79 Resection: 59/79	Favoring biopsy: 6.8 (4.8–9.6)2 Favoring resection: 6.9 (3.7–9.9)2	EuroQol-5D QLQ-C30 QLQ-BN20
Klein et al 200326	As in Aaronson et al 2011 plus: irradiated patients included only if received radiotherapy as primary treatment.	Biopsy: 72/156 Resection: 84/156 Radiotherapy: 86/156	N/A	SF-36
Reijneveld et al. 200166	Supratentorial DLGG Underwent biopsy or resection ≥6 mo previously, without progression since then. Age ≥18 y	Biopsy: 6/48 Surgery: 17/48	Suspected LGG: 4.4 (4.5) Proven LGG: 5.5 (3.7)	SF-36 QLQ-BN20 Cognitive Functioning Scale (CFS)
Pretreatment newly diagnosed
Ruge et al 201127	Imaging-suspected, newly diagnosed, untreated supratentorial DLGG.	None	Newly diagnosed	SF-36

Abbreviations: DLGG, diffuse hemispheric WHO grade II glioma; LGG, low-grade glioma; mo, month; y, year.

If N/A, information could not be ascertained from the manuscript or after requesting information from authors.

^aLater reading of subsample published in Aaronson et al 2011.

^bMedian (IQR).

Participation, Dropout Rates, and Missing Items

Level of participation varied with a weighted mean of 64.0% (range: 32.7%–95.3%, n = 16, Table 4). Twelve studies provided specific reasons with sample sizes for nonparticipation. Two studies gave reasons without sample sizes,^15,32 while 3 studies gave no reasons for nonparticipation in the study.^33–35 Study dropout rates in the longitudinal studies varied with a weighted mean of 2.1% per month (range: 0%–7.9% per month, n = 10; online supplementary material Supplementary material, Table S3). Dropout occurred primarily as a result of disease progression or administrative error, the former representing a major selection bias in the reporting of QoL in this population. Two studies evaluating the use of temozolomide recorded dropout due to side effects of the medication, another significant bias.^36,37 One study attempted to use the EORTC QLQ-BN20; however, after collecting data for patients, this was abandoned as patients did not want to be prospectively confronted with questions about side effects relating to treatment or tumor progression at the time of presentation.²⁷ Only one study provided details of strategies to reduce dropout, with authors suggesting that reporting HRQoL should be a mandatory requirement of all clinical trials of patients with DLGG and that compliance should be closely monitored.¹⁵

Table 4.

Level of participation in quality of life assessment where described

Study	Participation (%)	Reasons for nonparticipation
Aaronson et al 201132	195/239 (81.6%)	Perceived burden of data collection Not wanting to be confronted with illness history.
Boele et al 201540a	65/195 (33.3%)	Deceased: 58 (30%) Tumor recurrence/treatment: 45 (23%) Declined: 6 (3%) Other reasons: 19 (10%) Incomplete HRQoL data: 2 (1%)
Budrukkar et al 200967	243/257 (94.6%)	Data were not available for 14 patients due to administrative errors or inability to complete the questionnaire due to profound neurological deficits.
Correa et al 200839	25/40 (62.5%)	Disease progression: 8 Unavailable/declined: 7
Edvardsson et al 200933	39/49 (79.6%)	N/A
Gehring et al 200930	140/366 (38.2%)	No cognitive deficits/complaints (therefore ineligible for study): 132 Serious visual/verbal/motor/cognitive handicap: 25 No time: 18 Emotionally too confronting: 17 Received other psychological therapy: 6 Inaccessible: 6 Intervening medical problems: 5 Lack of motivation: 4 Psychiatric disorders: 2 Other: 11
Gustafsson et al 200638	38/49 (77.6%)	Declined: 10
Jakola et al 201128	88/197 (44.7%)	Nonresponders: 109
Jakola et al 201411	79/153 (51.6%)	Deceased: 68 Incapable of participating: 3 Nonresponders: 2 Refused: 1
Kiebert et al 199815	113/345 (32.7%)	Center nonparticipation major factor. Compliance good in patients who participated, but no data on individual reasons for nonparticipation available.
Klein et al 200326	156/239 (65.3%)	Too burdensome/reluctant to be confronted with “cured” illness: 44 (18%) Missing data: 39 (20%)
Liu et al 200937	62/66 (93.9%)	Not LGG: 1 Missing questionnaires: 3
Osman 201434	18/20 (90.0%)	N/A
Pace et al 200335	43/74 (58.1%)	N/A
Taphoorn et al 199468	41/43 (95.3%)	Serious visual/verbal/motor/cognitive handicap: 1 Psychiatric disorders: 1
Yu et al 201365	58/153 (37.9%)	Disagreed to participate: 26 Did not meet inclusion/exclusion criteria: 69

Abbreviations: HRQoL, health-related quality of life; LGG, low-grade glioma/

If N/A, information could not be ascertained from the manuscript or after requesting information from authors.

^aLater reading of subsample published in Aaronson et al 2011.

Four of the 26 studies referred to missing items in completed questionnaires (Table 5). Two studies provided evidence from validation studies to support their treatment of missing items.^15,37 Strategies varied from excluding forms with missing data,³⁸ imputing values if more than half of items were provided for a subscore,^11,15 imputing based on other questionnaires used concomitantly,¹¹ and calculating total scores if overall item response rate exceeded 80%.³⁷

Table 5.

Strategies to handle missing data

Study	HRQoL Measure	Missing Item Statistical Approach
Gustafsson et al 200638	EORTC QLQ-C30	Exclude if missing items
Jakola et al 201411	EORTC QLQ-BN20	Imputed the mean for that patient in missing item if >50% items present. In multi-item domains with >50% missing, they were kept as missing (n = 2, “future uncertainty,” and “global HRQoL”). Missing single items kept as missing.
	EORTC QLQ-C30
	EQ-5D	Imputed based on EORTC questionnaires and other EQ-5D domains.
Kiebert et al 199815	Customized questionnaire	For missing items within a scale, if >50% items completed, scale score calculated using the completed items.
Liu et al 200937	FACT-Br	Subscale calculation acceptable as long as >50% items completed, total score appropriate as long as overall item response rate >80%.

International Society for Quality of Life

Full ISOQOL results are provided in the online supplementary material (Supplementary material, Table S4), but some of the salient findings are described. Eighty-one percent of the studies mentioned the HRQoL measure as a predefined primary or secondary outcome measure, with all studies reporting a primary outcome measure including HRQoL in the title (n = 16). Two studies (8%) reporting HRQoL did not identify this in the abstract.^34,39 All but one study reporting HRQoL as a primary outcome provided an adequate summary of the literature (n = 15/16, 94%). Less than a third of studies provided a specific HRQoL hypothesis (n = 8/26, 31%), while no studies reporting HRQoL as a primary outcome provided rationale or specified expected directions of change or time points of assessment.

With regards to the methods, none of the 18 studies reporting the mode of HRQoL measuring administration to participants specified an electronic format. Every study reporting HRQoL as a primary outcome provided a citation of the original development of the instrument, which in all cases had been published previously. However, few studies provided a rationale for the choice of the HRQoL instrument (n = 12/26, 46%). None of the studies reporting HRQoL as a primary outcome provided a rationale for the window of responses. The majority of studies gave evidence of instrument validity and reliability (n = 14/26, 54%), although as mentioned previously, none of the measures had been validated in patients with DLGG specifically. The majority of studies with a longitudinal design published data collection schedules (n = 10/17, 59%).

Statistical analysis was generally well-reported, with the majority of studies performing appropriate tests (n = 21/26, 81%) and providing data on patients' baseline characteristics (n = 23/26, 88%). Many described the extent of missing data (n = 21/26, 86%), but only a minority of studies provided statistical methods to deal with the missing data in analysis (n = 5/21, 24%). There was one prospective study with HRQoL as a primary outcome that did not compute a power-to-sample size calculation based on the HRQoL instrument used.²⁹ Studies were exposed to selection bias as only a minority of relevant studies adjusted for survival differences between treatment groups (n = 2/8, 25%).

While all studies contextualized the results with clinical implications, very few specifically defined thresholds for clinical significance of findings from HRQoL data (n = 5/26, 19%). Of these, only 2 studies provided a methodology to support the described thresholds.^37,40 Four studies reporting HRQoL as a primary outcome used predefined responder definitions, all describing the proportions of patients successfully achieving them.^28,32,37,40

Discussion

Discussion of Results

This systematic review examining the reporting of QoL in patients with DLGG identified 11 different HRQoL measures from 26 studies. The EORTC QLQ-BN20 was the most commonly used measure and was identified in 11 accepted studies. Despite the widespread use of these measures and evidence of reliability and validity across 12 different languages, there is no existing evidence of validation of their use in patients with DLGG. The role of QoL and its assessment are very different in DLGG relative to patients with glioblastoma or other brain tumors due to key differences in average age of presentation, treatment, clinical course of the disease, and ultimately prognosis. It may be the case that metrics specific to DLGG patients need to be developed and validated in order to capture salient aspects of the HRQoL in these patients that may either be missed (eg, cognitive functioning, epilepsy severity) or over-represented (eg, hair loss) by the HRQoL scales currently in use for this population.

While multiple studies applied the same HRQoL measure, heterogeneity in inclusion criteria, the interventions used, the comparators used, and the data collection schedule all limited the potential for pooled analysis. The limited evidence suggested preservation of HRQoL following temozolomide chemotherapy but a slow decline in QoL with long-term follow-up over a period of 24 months (Supplementary Material, Fig. S2). Analysis of the EORTC QLQ-BN20 and SF-36 HRQoL metrics in long-term follow-up suggested that HRQoL was preserved across a wide spectrum of posttreatment follow-up periods, with significantly greater variation between participants within studies than across the studies included (Supplementary Material, Figs S3 and S4). Role-functioning deteriorated over time, despite relative preservation of physical and mental health domains. In addition to highlighting important domains that may not be assessed by objective or physician-assessed outcome measures, these data identified an expressed need to ascertain the reasons behind the functional decline in DLGG patients despite relatively preserved physical and mental health.

Low rates of participation (mean participation rate: 64.0%, range: 32.7%–95.3%) and high rates of dropout (weighted mean dropout rate: 2.1% per month) exposed the published data to selection bias. Although several studies conducted statistical tests to identify differences between participants and nonparticipants, this was not universal. While dropout rates were generally well-reported, strategies to reduce the dropout were limited. The reporting of disease progression as a cause for nonparticipation and dropout is problematic. The continued collection of HRQoL data is important to minimize selection bias⁴¹ and ensure accurate profiling of longitudinal HRQoL for patients with DLGG.

The ISOQOL reporting standards are a pivotal tool that could ensure appropriate presentation of data in published studies of HRQoL in patients with DLGG.²² The items were adapted in this study (excluding studies from items not relevant to their design) to enable fair presentation of the quality of reporting across study types. Analysis of the levels of reporting across items demonstrated a lack of adequate HRQoL hypothesis generation in combination with limited evidence to support the design of assessment schedules. The absence of statistical methods to account for the low levels of reported participation and high levels of reported dropout and differing survival rates in study arms expose the published data to selection biases. None of the studies leveraged electronic methods to collect HRQoL data. The use of HRQoL instruments not validated in the DLGG population was compounded by the absence of robust methodologies to determine clinical significance of improvements or declines observed.

Limitations of the Current Study

Included studies frequently did not specify patients with DLGG or provide a subgroup analysis in the published manuscript, which reflected an underappreciation of the differences in pathology, treatment, and disease progression between DLGG and other brain tumors. Even within the DLGG cohort, there may be differences in the pathology, treatment, and prognosis for different WHO grade II tumors; advances in molecular/genetic classifications, complemented by a standardized schedule of HRQoL outcomes, may elicit important differences in the prognosis and QoL outcomes for these different subgroups.^42,43 Availability of individual patient-level data using a standardized schedule of HRQoL assessments is recommended in future studies of HRQoL in DLGG to enable robust assessment and comparison of patient outcomes across studies.

Future Directions

The development of standards for HRQoL data collection specific to DLGG is recommended in order to enable such analyses to capitalize on improved sample sizes and facilitate cross-study comparisons. Such standards should include a mandatory data collection schedule at all stages in the disease, particularly at baseline (pretreatment), posttreatment, and standard intervals of follow-up. With more consistent reporting of HRQoL, patient-level data could be acquired, and the proportion of patients recording a clinically significant change in HRQoL can be assessed.^40,44 Standardized schedules will also be crucial to ascertain the different temporal profiles of changes in HRQoL brought about by different treatments. It will also help highlight differences in HRQoL for patients with DLGG in different anatomical locations or with different molecular/genetic subtypes.

In order to optimize participation, dropout, and missing items on forms, technology could be leveraged to program a schedule for data collection in longitudinal studies with the provision of electronic forms and automated timely reminders at standardized intervals; HRQoL data collection with electronic forms has already been achieved successfully using a tablet computer.^45,46 HRQoL measurements should be complemented alongside a thorough evaluation of neurocognitive functioning and epilepsy severity, both of which represent core outcome measures in patients with DLGG.⁵ Incorporation of these assessments in addition to traditional objective measures such as overall survival and progression-free survival will help determine the net clinical benefit of various treatment strategies.⁴⁷ Furthermore, consideration of the ISOQOL minimum standards for patient-reported HRQoL measures (including adequate reliability, validity, and responsiveness) is recommended, although it is recognized that these have been designed specifically for RCTs and would therefore need to be adapted for other study designs.⁴⁸

Finally, the process of developing a standardized assessment schedule and set of HRQoL metrics should follow a robust, established methodology similar to that employed by the Core Outcomes Measures in Effectiveness Trials (COMET) initiative when developing core outcome sets for specific conditions.⁴⁹ In particular, this consensus process should include all relevant stakeholders including neurosurgeons, neuro-oncologists, academics, QoL experts, patients, and carers.

Undoubtedly, there will be debate as to the specific HRQoL instrument(s) used and whether these will be generic or disease-specific instruments, a debate that has been discussed in a number of different fields including traumatic brain injury.⁵⁰ The advantages of generic scales include comparison across diseases and a healthy population, whereas disease-specific scales are likely to better cover the salient domains, which in the case of DLGG may include domains such as epilepsy, cognitive, and role functioning. Regardless, the chosen instrument(s) require validation in DLGG cohorts and assessments for sensitivity and minimal clinically significant differences.

Conclusion

The pursuit of a robust and comparable assessment of QoL across randomized and nonrandomized studies is of utmost importance in this population as therapies continue to develop that may prolong survival at the cost of side effects and treatment-associated neurological deficits. This study has presented the current status of reporting patient-reported health-related QoL in patients with diffuse low-grade glioma and serves as a basis for developing a standardized schedule and set of QoL measures for robust study design and data collection in the future.

Supplementary Material

Supplementary material is available online at Neuro-Oncology (http://neuro-oncology.oxfordjournals.org/).