Measurement of quality of life in second-line patients with advanced NSCLC without targetable mutations: a review

Abstract

Quality of life (QoL) is important to cancer patients and is increasingly included as a trial end point. The methodologies/findings of randomized controlled trials evaluating the efficacy and safety of second-line treatments approved for use in the EU in patients with advanced/metastatic NSCLC, without known targetable mutations, were evaluated. Seven trials were identified; five compared active treatments and two compared active treatment to placebo. Methodologies used and reporting varied. The European Organization for Research and Treatment of Cancer lung cancer questionnaire was the most commonly used assessment method (n = 4). There was no evidence to suggest differences in QoL between active treatments. Consistent and appropriate use of standard QoL instruments in future would increase the reliability of results and their applicability to clinical decision-making.

Practice points.

• This article provides a comprehensive review of controlled clinical trials evaluating the impact of approved treatments for NSCLC after first-line therapy on quality of life (QoL) and evaluation of their quality based on important factors for adequate evaluation of QoL.

• QoL issues are important in advanced NSCLC and are especially important the more advanced the treatment line.

• Clinicians should consider all available data when making treatment decisions and QoL data should be included in clinical decision-making.

• When evaluating QoL outcomes of a treatment, it is essential that the instrument being used is suitable for the patient population being examined and the goals of treatment.

Lung cancer is the most commonly diagnosed cancer and the leading cause of cancer mortality in the world [1], with NSCLC accounting for more than 85% of all cases [2,3]. Treatment of NSCLC is guided by both histology and the presence of targetable mutations. Patients with advanced and/or metastatic NSCLC of adenocarcinoma histology, without known EGFR mutations or ALK rearrangements, usually receive a platinum-based regimen in the first-line setting [2,4,5]. However, nearly all patients will eventually relapse, and require further lines of treatment. In the EU, until recently, second-line treatment options for these patients were limited to docetaxel, pemetrexed and erlotinib; however, nintedanib in combination with docetaxel is also now approved for the treatment of patients with locally advanced, metastatic or locally recurrent NSCLC of adenocarcinoma histology after first-line chemotherapy.

Although extending survival is one of the most important goals of cancer treatment, improvements or maintenance of quality of life (QoL) are also of particular importance to patients who require palliative care. QoL measurements help to include a human perspective in oncology treatment and patients appreciate evaluation of their overall wellbeing by additional QoL questionnaires [6]. Over the last decade, the evaluation of patient-reported outcomes (PRO) data, such as QoL assessment, has increased as a secondary outcome in oncology studies, whereas the primary outcome has remained focused on overall survival (OS) or progression-free survival (PFS). Given that these studies are powered to detect differences in the primary outcome, the value of the QoL rating is often only descriptive. To increase the validity of such results, it is important that trials including QoL measures have a valid research question and are appropriately designed.

A number of standard instruments have been developed and validated to measure the impact of a given treatment on QoL. Evaluation of QoL is complex given that it is multidimensional and incorporates elements of physical, psychological and social functioning, and wellbeing. Several QoL instruments are available specifically for use in patients with cancer. The European Organization for Research and Treatment of Cancer (EORTC) multidimensional core questionnaire QLQ-C30 (EORTC QLQ-C30) [7] for Europe and the Functional Assessment of Cancer Therapy – General (FACT-G) [8] questionnaire for North America are both standard and commonly used instruments that have been shown to have excellent psychometric characteristics. Both questionnaires also include validated lung cancer-specific modules. Details of the QoL questionnaires are given in Table 1. General QoL questionnaires also exist, such as the Short Form (36) Health Survey, a 36-item, patient-reported survey of patient health [9].

Table 1. . Summary of different general and lung cancer-specific quality of life questionnaires.

Details	Questionnaires
	EORTC QLQ-C30 [7]	EORTC QLQ-LC13 [10]	FACT-L [11]	LCSS [12]
Number of questions	30	13	36	15
Components	Five functional scales: – Physical (subjective feelings of physical health) – Role (subjective feelings of ability to work) – Cognitive (subjective feelings of cognitive functioning and concentration) – Emotional (emotional wellbeing) – Social Three symptom scales: – Fatigue – Pain – Nausea/vomiting A global health status/QoL scale Individual items: – Dyspnea – Insomnia – Appetite loss – Constipation – Diarrhea – Financial difficulties	Disease-specific modular questionnaire that should always be complemented by the QLQ-C30 Two domains/categories: – Lung cancer-related symptoms – Treatment side effects	Five domains/categories: – Physical – Social/family – Emotional – Functional wellbeing – Lung cancer subscale (symptoms, cognitive function, regret of smoking)	Patient scale domains of nine items related to: – Symptoms – Total symptomatic distress – Activity status – Overall QoL Observer scale: symptoms includes a subscore using the mean of all six major symptoms (ASBI)
Scoring	Items 1–28: 4-point scale from 1 (not at all) to 4 (very much); items 29 + 30: seven-point scale from 1 (very poor) to 7 (excellent)	4-point scale from 1 (not at all) to 4 (very much)	5-point scale from 0 (not at all) to 4 (very much)	Patient scale: 100 mm horizontal line (0 = lowest rating; 100 = highest rating) Observer scale: 5-point categorical scale (100 = none; 0 = severe)
Time to complete	11–12 min	<10 min	11–12 min	10–15 min

ASBI: Average symptom burden index; EORTC: European Organization for Research and Treatment of Cancer; EORTC QLQ-C30: EORTC multidimensional core questionnaire; EORTC QLQ-LC13: EORTC lung cancer questionnaire; FACT-L: Functional Assessment of Cancer Therapy – Lung; LCSS: Lung Cancer Symptom Scale; QoL: Quality of life.

Regardless of the stage of their disease, all patients with cancer are concerned about their QoL under treatment. The concern is particularly high for patients with advanced disease, where the OS prognosis may be limited to months rather than years. As such, adequate assessment of QoL is of particular importance to complement the conventional end points in oncology studies, especially in the palliative situation. We aimed to review QoL data collected in controlled studies, evaluating the use of approved treatments for NSCLC after first-line therapy and to evaluate their quality based on several factors for adequate evaluation of QoL: a defined clinical question with appropriate hypothesis, methodology and study design; use of reliable, appropriate and validated questionnaires; and adequate documentation of the methods for QoL assessment.

Methods

• Literature identification

A literature review of PubMed was undertaken to identify all relevant randomized controlled trials evaluating the efficacy and safety of treatments for advanced or metastatic NSCLC after first-line therapy. Studies identified by the review were analyzed according to the following inclusion criteria:

• Adults with histologically or cytologically confirmed, locally advanced and/or metastatic NSCLC of stage IIIB or IV or recurrent NSCLC (all histologies);

• Prior therapy: previous first-line therapy with monotherapy or combination chemotherapy;

• Study design: randomized, controlled Phase III trials;

• Outcomes: any QoL evaluation;

• Interventions: studies investigating treatments approved for use in the EU for the second-line setting in patients with NSCLC without known mutations.

Approved treatments included erlotinib, pemetrexed, docetaxel, nivolumab and nintedanib in combination with docetaxel. EMBASE was also searched to identify congress presentations of QoL data for trials that did not report QoL outcomes in the primary publication. Searches were limited to English language publications. Exclusion criteria included documented EGFR mutation-positive NSCLC or other mutations with recognized treatment targets; patients who were treatment-naive or who had received more than first-line therapy; and studies including non-licensed treatments as either the experimental or control arm.

• Data extraction

All identified literature was reviewed and QoL data were extracted. All pertinent information regarding QoL measurement was captured, including details of rating scales used, frequency of assessment and statistical methods for analysis (for example, time-to-deterioration analysis, mean treatment difference). Documentation of results included compliance with QoL evaluations, baseline scores (where reported), overall QoL effects (measured using the EORTC QLQ-LC13, EORTC QLQ-C30, LCSS or the FACT-L) and NSCLC disease-specific effects related to key symptoms of pain, cough, dyspnea and fatigue.

Results

• Results of literature search: identified trials

The search of the published literature as of 30 September 2015 identified 161 citations that were initially reviewed and screened, resulting in evaluation of 44 full-text articles. Of the papers identified, 14 publications reporting QoL results from seven trials in patients treated with EU-approved second-line treatment options for NSCLC of adenocarcinoma histology were identified. The patient characteristics of the identified studies are presented in Table 2. All but one study included patients with one previous treatment; BR.21 enrolled patients who had received one or two lines of previous treatment. Five studies compared different active treatments, whereas two trials compared active treatment to placebo/best supportive care. One trial (TAILOR) has not yet reported the QoL findings or details of their methodologies. Most studies included patients with NSCLC of any histology, although CheckMate 017 was conducted in patients with squamous cell histology only [13].

Table 2. . Design and patient characteristics of identified studies.

Study	Patients	Previous treatment	Treatments		Efficacy and safety outcomes			QoL assessment			Ref.
			Experimental arm	Control arm	OS (months)	PFS (months)	Tolerability (most common adverse events)	Scale	Frequency of assessment	Outcomes
LUME-Lung 1	Unselected stage IIIB/IV NSCLC (all histologies)	Chemotherapy	Docetaxel + nintedanib	Docetaxel + placebo	12.6 vs 10.3; p = 0.0359 (key secondary end point in adenocarcinoma patients)	3.4 vs 2.7; p = 0.0019	Grade ≥3: diarrhea (3.6 vs 2.6%); reversible increased ALT (7.8 vs 0.9%; reversible increased ALT (3.4 vs 0.5%)	EORTC QLQ-LC13, EORTC QLQ-C30, EQ-5D and EQ-VAS	Baseline, end of every cycle, and end of therapy and at the first follow-up visit	TTD (first appearance of a minimal clinically important difference [≥10-point change]) and longitudinal assessment of cough, dyspnea and pain	[4,17]
CheckMate 017	Stage IIIB/IV NSCLC (squamous cell histology)	Prior systemic therapies	Nivolumab	Docetaxel	9.2 vs 6.0; p < 0.001 (primary end point)	3.5 vs 2.8; p < 0.001	Grade 3–4: neutropenia (0 vs 30%)	LCSS (ASBI and 3-Item Index), EQ-5D and EQ-VAS	Baseline, every 4 weeks (nivolumab arm) and every 3 weeks (docetaxel arm) for first 6 months	TTD, mean change in scores over time, LCSS ASBI improvement (≥10-point change) by week 12	[13,16]
BR.21	Unselected NSCLC (all histologies)	Chemotherapy (one or two lines)	Erlotinib	Placebo	6.7 vs 4.7; p < 0.001 (primary end point)	2.2 vs 1.8; p < 0.001	All grades: rash (76 vs 17%); anorexia (69 vs 56%); stomatitis (19 vs 3%)	EORTC QLQ-LC13, EORTC QLQ-C30	Baseline, every 4 weeks during treatment, 4 weeks after completing treatment and every 12 weeks thereafter until documentation of PD	TTD (first appearance of a minimal clinically important difference [≥10-point change]), cough, dyspnea and pain (primary outcome for QoL) Percentage improved/stable/worse	[18,19]
JMEI	Advanced NSCLC (all histologies)	Chemotherapy	Pemetrexed	Docetaxel	8.3 vs 7.9; p = NS (primary end point)	2.9 vs 2.9	Grade 3–4: neutropenia (5 vs 40%); febrile neutropenia (2 vs 13%)	LCSS (ASBI)	NR	Percentage of patients who were improved, stable or worse (or meaningful change) on ASBI (>0.5 SD change from baseline) or LCSS (1-point change)	[20,21]
TITAN	Advanced NSCLC (all histologies)	Platinum-based therapy	Erlotinib	Docetaxel/pemetrexed	5.3 vs 5.5; p = 0.73 (primary end point)	6.3 vs 8.6; p = 0.089	All grades: rash (50 vs 5%); diarrhea (18 vs 2%); alopecia (0 vs 11%)	FACT-L, version 4	Baseline, every 3 weeks until Week 48 and every 12 weeks thereafter until PD	TTP TTD	[22]
TAILOR	Advanced, EGFR wild-type NSCLC (all histologies)	Platinum-based therapy	Erlotinib	Docetaxel	5.4 vs 8.2; p = 0.05 (primary end point)	2.4 vs 2.9; p = 0.02	Grade 3–4: low absolute neutrophil count (0 vs 20%); skin toxic effects (14 vs 0%) and asthenia (6 vs 10%)	EORTC QLQ-LC13, EORTC QLQ-C30	Baseline and before each treatment cycle	–	[14,15]
TAX-317	Advanced IIIB/IV NSCLC (all histologies)	Chemotherapy	Docetaxel	Best supportive care	7.0 vs 4.6; p = 0.047 (primary end point)	TTP (weeks): 10.6 vs 6.7; p < 0.001	Grade 3 or 4 neutropenia: 86% (100 mg/m2) and 67% (75 mg/m2)	LCSS EORTC QLQ-LC13	Baseline, immediately before each treatment cycle, at the end of drug treatment and every 2 months during follow-up	Longitudinal analysis; mixed-model analysis and ANCOVA	[23,24]

ANCOVA: Analysis of covariance; ASBI: Average symptom burden index; EORTC: European Organization for Research and Treatment of Cancer; EORTC QLQ-C30: EORTC multidimensional core questionnaire; EORTC QLQ-LC13: EORTC lung cancer questionnaire; EQ-5D: EuroQol-5D; EQ-VAS: EuroQol-visual analog scale; FACT-L: Functional Assessment of Cancer Therapy - Lung; LCSS: Lung Cancer Symptom Scale; NR: Not reported; NS: Not significant; OS: Overall survival; PD: Progressive disease; PFS: Progression-free survival; QoL: Quality of life; RR: Relative risk; SD: Standard deviation; TTD: Time to deterioration; TTP: Time to progression.

• Study design & methodology in identified trials

QoL instruments used in identified trials & frequency of assessment

Trials used a variety of lung cancer-specific QoL instruments (Table 2). EORTC QLQ-LC13 was the most commonly used assessment method, used in four studies, followed by the EORTC QLQ-C30 and the LCSS, used in three studies each. The FACT-L was used the least commonly (one study). All trials conducted assessments at baseline and at regular intervals corresponding to the frequency of treatment cycles. Only one trial (LUME-Lung 1) specifically reported that questionnaires were completed by patients before seeing the investigator, and before they were provided with any new information about their disease status, to avoid bias.

QoL methodologies

All trials used OS or PFS as the primary end point, with QoL measures being included as secondary end points. Methodologies for the evaluation of QoL changes varied between studies, with no one method being consistently used across trials. Methods used included time-to-deterioration in symptoms (n = 5), longitudinal assessment of symptoms (n = 3) and categorical assessment of the percentage of patients with improvement or worsening of symptoms (n = 3). One trial (TAILOR) did not report further details of assessment methodology [14,15] and one trial (CheckMate 017) has not been published in full [13,16]. Only two studies (LUME-Lung 1 and TAX-317) reported details of how missing data was accounted for in the statistical analysis, with both studies conducting analyses to evaluate the impact of missing data on the study findings. In LUME-Lung 1, missing data was not shown to impact on the interpretation of study findings, whereas patients completing final PRO assessments in TAX-317 had better improvements in pain than those not completing final PRO assessments.

Completion of QoL questionnaires & baseline scores

Reporting of compliance with questionnaire completion was good; all seven trials reported details of questionnaire completion. Table 3 shows that the methods of reporting and levels of questionnaire completion did vary between studies. Across all trials, completion rates decreased over the duration of the trial. Baseline QoL scores in each of the studies are shown in Table 4. Baseline values indicate that patients generally had QoL reflecting the performance status of the patient population typically included in clinical trials.

Table 3. . Quality of life assessment completion rates.

Study	Treatment	QoL assessment completion rates (%)		Ref.
		Baseline	During treatment
LUME-Lung 1	Docetaxel + placebo Docetaxel + nintedanib	>85	˜80 (during treatment)	[4,17]
CheckMate 017	Nivolumab Docetaxel	78 77	68 (week 12; LCSS) 67 (week 12; LCSS)	[13,16]
BR.21	Erlotinib Placebo	93.0 93.8	58.4 (at progression) 55.1 (at progression)	[18,19]
JMEI	Docetaxel Pemetrexed	>80 (based on evaluable patients only)		[20,21]
TITAN	Docetaxel/pemetrexed Erlotinib	˜90	˜80 (during treatment)	[22]
TAX-317	Docetaxel Best supportive care	75 (baseline and at least on one treatment assessment) 68 (baseline and at least on one treatment assessment)		[23]

LCSS: Lung Cancer Symptom Scale; QoL: Quality of life.

Table 4. . Baseline quality of life scores.

Study	Treatment	Scale	Baseline global QoL	Fatigue	Pain	Cough	Dyspnea	Ref.
LUME-Lung 1	Docetaxel + placebo Docetaxel + nintedanib	EORTC QLQ-LC13 and EORTC QLQ-C30	62.3 61.2	–	27.6 27.0	35.9 39.6	28.3 29.8	[4,17]
CheckMate 017	Nivolumab Docetaxel	LCSS ASBI	29.6 29.6	–	–	–	–	[13,16]
BR.21	Erlotinib Placebo	EORTC QLQ-LC13 and EORTC QLQ-C30	55.3 53.5	42.5 45.4	34.2 38.3	43.4 39.0	31.9 33.5	[18,19]
TAX-317	Docetaxel Best supportive care	LCSS	64.6 66.8	63.5 62.8	74.7 76.1	70.3 66.2	63.7 70.4	[23]

ASBI: Average Symptom Burden Index; EORTC: European Organization for Research and Treatment of Cancer; EORTC QLQ-C30: EORTC multidimensional core questionnaire; EORTC QLQ-LC13: EORTC lung cancer questionnaire; LCSS: Lung Cancer Symptom Scale; QoL: Quality of life.

• PRO results from identified trials

Overall QoL ratings

All trials reported on the impact of treatment on overall QoL, although no studies comparing active treatments reported significantly different improvements in QoL. Three studies evaluated changes in overall QoL using median time-to-deterioration analysis (LUME-Lung 1, CheckMate 017 and TITAN) (Table 5) and three studies (LUME-Lung 1, CheckMate 017 and TAX-317) also evaluated overall QoL improvements using longitudinal assessment (Table 6). A further three trials (JMEI, CheckMate 017 and BR.21) evaluated QoL by assessing the proportion of patients with symptom improvement (Table 7). Only one of the two placebo-controlled studies showed a significant difference in overall QoL; erlotinib was shown to be associated with significant improvement in overall QoL compared with placebo, as shown by the percentage of patients with improvement in EORTC QLQ-C30 Global QoL scores. Longitudinal analysis of EORTC QLQ-C30 Global QoL scores in TAX-317 showed a trend towards improvement with docetaxel compared with best supportive care but the difference was not statistically significant. Effects of treatment on role function and physical function were not routinely reported, although two trials did report differences in mean physical function scores by longitudinal assessment: LUME-Lung 1 reported a trend favoring nintedanib in combination with docetaxel over placebo in combination with docetaxel in patients with adenocarcinoma and TAX-317 reported a trend favoring docetaxel over best supportive care.

Table 5. . Median time-to-deterioration of quality of life or symptoms.

Study	Treatment	Global QoL, HR (95% CI)	Fatigue, HR (95% CI)	Pain, HR (95% CI)	Cough, HR (95% CI)	Dyspnea, HR (95% CI)	Ref.
LUME-Lung 1 All patients	Docetaxel + placebo Docetaxel + nintedanib	0.95 (0.83–1.10)	–	0.95 (0.82–1.09)	0.90 (0.77–1.05)	1.05 (0.91–1.20)	[4,17]
LUME-Lung 1 Adenocarcinoma patients	Docetaxel + placebo Docetaxel + nintedanib	0.86 (0.71–1.05)	0.97 (0.79–1.19)	0.93 (0.76–1.14)	0.97 (0.78–1.20)	1.04 (0.86–1.26)	[4,17]
CheckMate 017	Nivolumab Docetaxel	0.67 (0.43–1.03) LCSS ASBI	–	–	–	–	[13,16]
TITAN	Docetaxel/pemetrexed Erlotinib	TTP: HR: 1.19 (95% CI: 0.90–1.57; p = 0.22) TTD: HR: 1.21 (95% CI: 0.93–1.59; p = 0.15)	–		–	–	[22]
BR.21	Erlotinib Placebo	–	–	2.8 months (95% CI: 2.4–3.0) and 1.9 months (95% CI: 1.8–2.8), respectively (p < 0.03)	4.9 months (95% CI: 3.8–7.4) for patients receiving erlotinib, and 3.7 months (95% CI: 2.0–4.9) for patients receiving placebo (p < 0.04)	4.7 months (95% CI: 3.8–6.2) and 2.9 months (95% CI: 2.0–4.8), respectively; (p < 0.04)	[18,19]

ASBI: Average Symptom Burden Index; HR: Hazard ratio; QoL: Quality of life; TTD: Time to deterioration; TTP: Time to progression.

Table 6. . Longitudinal model/mean change in quality of life or symptom scores.

Study	Treatment	Global QoL	Fatigue	Pain	Cough	Dyspnea	Ref.
LUME-Lung 1 Adenocarcinoma patients	Docetaxel + placebo Docetaxel + nintedanib	Trend favoring nintedanib	–	Trend favoring nintedanib	Trend favoring nintedanib	NS	[4,17]
CheckMate 017	Nivolumab Docetaxel	Trend favoring nivolumab on LCSS and ASBI	–	–	–	–	[13,16]
TAX-317	Docetaxel (overall) Best supportive care	Trend favoring docetaxel only	NS	Favored docetaxel	NS	NS	[23]

ASBI: Average symptom burden index; LCSS, Lung Cancer Symptom Scale; NS: Not significant; QoL: Quality of life.

Table 7. . Proportion of patients with improved/stable/worse quality of life or symptom scores.

Study	Treatment	Global QoL	Fatigue	Pain	Cough	Dyspnea	Ref.
JMEI	Docetaxel Pemetrexed	No significant difference (ASBI)	Improved/stable: 54.8 vs 56.7%	Improved/stable: 64.0 vs 62.1%	Improved/stable: 63.6 vs 64.4%	Improved/stable: 63.6 vs 59.9%	[20,21]
CheckMate 017	Nivolumab Docetaxel	LCSS improvement at week 12: 20 vs 21.9%	–	–	–	–	[13,16]
BR.21	Erlotinib Placebo	Improved: 35 vs 26%; p < 0.0001	Improved: 45 vs 36%; NS	Improved: 42 vs 28%; p < 0.01	Improved: 44 vs 27%; p < 0.0001	Improved: 34 vs 23%; p = 0.03	[18,19]

ASBI: Average symptom burden index; LCSS: Lung Cancer Symptom Scale; NS: Not significant; QoL: Quality of life.

Symptom ratings

Results for the NSCLC symptoms of interest defined in this review (fatigue, pain, cough and dyspnea) were reported in four out of five studies reporting results (Tables 5–7), although the analysis methods differed between studies. No significant differences between treatment arms with respect to these symptoms were reported in studies using active controls. Significant results were reported for erlotinib when compared with placebo for pain, cough and dyspnea, although improvements in fatigue did not reach significance. Only one study, LUME-Lung 1, reported findings separately for patients with adenocarcinoma histology NSCLC. Although the CheckMate 017 study did not report results for the NSCLC symptoms of interest defined in this review, results were reported for time to first disease-related deterioration on the LCSS 3-item index – a composite of symptom distress, interference with activity levels and health-related QoL – and showed an improvement with nivolumab compared with docetaxel (hazard ratio [HR]: 0.57; 95% CI: 0.38–0.85).

Discussion

Our review of QoL measured by PRO data in patients with advanced NSCLC without targetable mutations undergoing second-line treatment evaluated data from seven trials investigating four treatment options approved in the EU for use in patients with NSCLC. The trials showed a diverse range of scales used to measure QoL and differed in analysis methods and reporting. In agreement with a previous review on this topic [25], statistically significant improvements in QoL are not generally seen in this treatment setting when comparing active treatments. Although comparison of results from different trials should always be conducted with caution due to inherent differences in trial methodologies, the heterogeneity of QoL data in the assessment makes cross-trial comparisons impossible. As such, evaluation of the quality of the data reported is an important component of evaluating the value of data to clinical decision-making.

As stated in the introduction, adequate evaluation of QoL depends on several factors – first and foremost is a defined question and appropriate study design. It is important that all clinical studies are designed around a genuine question of interest. In the case of the NSCLC studies included in this review, the primary focus of the studies (the primary end point) was to establish whether the experimental interventions significantly improved either PFS or OS. As required by the EMA, QoL assessments are often included in clinical studies as one of the secondary objectives in order to complement the efficacy results of the primary end point [26]. As such, none of the studies reviewed here were designed around a QoL-related outcome or powered to detect significant differences in QoL. Hence, interpreting QoL findings can be challenging due to inherently wide confidence intervals of the results. One pragmatic solution that has been suggested is to lower the required significance level from 5 to 10% [27]; however, using this alternative threshold for significance would not have changed the conclusions of the majority of studies included in this review. Another approach to establishing significance is to define the ‘minimum clinically important difference’ or the magnitude of improvement that is considered to be clinically significant [28]. It is important that this improvement is defined a priori to study initiation. Four studies identified in this review reported QoL improvements using predefined criteria for improvement. JMEI defined meaningful change as at least a 1-point change on the 5-point LCSS or at least half a standard deviation change from baseline on the ASBI; CheckMate 017 used a ≥10-point change from baseline on the ASBI, and LUME-Lung 1 and BR.21 defined improvement as a 10-point change on the 100-point EORTC QLQ-C30.

Adequate QoL assessment also relies on the use of reliable, appropriate and validated instruments of measurement. As mentioned earlier, the EORTC QLQ-C30 and the FACT questionnaire are both standard, diagnosis-specific instruments for NSCLC research and are widely used. Indeed, the EORTC lung cancer-specific QLQ-LC13 was the most commonly used questionnaire identified in this review and was used in conjunction with the EORTC QLQ-C30 in three of the four studies.

In order to accurately evaluate the QoL findings from a trial, it is essential to have adequate documentation of the methods for QoL evaluation, assessment and data collection. Three trials identified in this review (BR.21, LUME-Lung 1 and TAX-317) had full publications dedicated to reporting the full QoL methodologies and results of the trial. Other studies included QoL outcomes in the primary publications (JMEI and TITAN), or have not yet been reported in full (CheckMate 017), making it difficult to fully evaluate the methodologies used. Despite this, all studies, with the exception of JMEI, reported details of the frequency of assessment.

When selecting an appropriate instrument for measuring QoL, it is important to consider the time taken to complete the questionnaire and the frequency of measurement over the time of the study conduct in order to keep the burden of assessment feasible for both the physician and patient and to get reliable QoL data. This is important for ethical but also methodological reasons: the higher the burden, the greater the probability of missing data. Missing data is also important when assessing the value of new treatments and, for this reason, the Joint Federal Committee (G-BA) who are responsible for benefit assessment of pharmaceuticals in Germany only accept data with a completion rate of >80; the Institute for Quality and Efficiency in Health Care uses a threshold of 70%. All studies identified in this review conducted assessments at similar frequencies (every 3–4 weeks), although questionnaire completion rates varied between trials, being highest for TITAN and LUME-Lung 1 (˜80% during treatment) and lowest for BR.21 (<60% at progression).

Timing of assessments is also important, especially when evaluating QoL in patients receiving chemotherapy. The assessment findings can be heavily influenced by when the patients received their treatment – before, during or after the QoL evaluation. All studies reviewed here evaluated QoL at the end of each chemotherapy treatment cycle. The relevance of timing may vary depending on the nature of the treatments being compared. For example, in the TITAN trial comparing chemotherapy to erlotinib, a potential bias in favor of erlotinib needs to be taken into account, as QoL assessment was conducted 20 days after the administration of chemotherapy and may not accurately capture the impact of chemotherapy on patients’ QoL. Conversely, in the LUME-Lung 1 trial, this potential bias does not need to be considered as chemotherapy was administered in both treatment groups.

In regard to assessment of factors leading to potential bias, it is of particular importance that study protocols include details of the questionnaire completion procedure: who hands out the forms to the patient, what kind of instructions are provided and at which time point during the study visit (e.g., prior to any CT scan evaluating tumor response). According to our experience, this type of information is frequently not reported but has important implications for assessing the quality of the data as these factors can substantially influence the collection of subjective QoL data. Only one trial identified in this review (LUME-Lung 1) provided this level of information and reported that questionnaires were completed by patients before seeing the investigator, and before they were provided with any new information about their disease status, to avoid influencing responses.

In addition to adequate documentation of assessments, the choice of adequate methods of analysis is important. The particular clinical question of interest needs to be defined and evaluations need to be prespecified accordingly to allow an appropriate assessment of the outcome. This is especially important when a large number of PRO items are evaluated, as the probability of a significant finding arising by chance increases as the number of variables measured increase. As such, it is important for the study protocol to consider what QoL domains are expected to be influenced by the treatment and to define by what extent improvement is expected (e.g., physical functioning is expected to improve by 15 points). Two studies identified in this review specifically defined symptoms of interest (LUME-Lung 1 and BR.21). Both trials stated that cough, dyspnea and pain were identified as clinically relevant and disease-specific symptoms of interest in the study analysis plans. Other trials reported results for these symptoms but did not indicate whether they were prespecified as symptoms of particular interest.

Selection of appropriate statistical analysis methods for QoL depends on a number of factors already discussed: trial design, context, and the specific research question of the trial and the timing of assessments. General considerations in selecting an appropriate methodology include variance and regression-analytic methods as standard techniques (parametric or nonparametric); multilevel (analyses and causal) indicator models might be the procedures of choice [23]. A number of different statistical analysis methodologies were used in the trials identified here. Although full discussion of the merits of each is beyond the scope of this paper, statistical handling of missing trial data is an important consideration when interpreting QoL findings. From our point of view, both the extent of missing data and how it is accounted for needs to be reported because, even if a study reports good completion rates (>80%), assumptions about the missing 20% of responses needs to be made (e.g., are they missing at random or are they missing because the patient is too weak or unwell to complete the questionnaire?). Questionnaire completion rates were generally reported in the studies identified in this review, and varied, whereas handling of missing data was not routinely reported. The two trials reporting the impact of missing data also showed differing results – LUME-Lung 1 showed no impact of missing data, whereas TAX-317 showed some effect.

Statistically significant improvements in global QoL measures were only observed in the two trials where active treatments were compared with placebo and were not reported in studies comparing active treatments directly. Both erlotinib and docetaxel appear to result in improvements in QoL when compared with placebo treatment; this is a positive finding given that both treatments are associated with adverse events (see Table 2) that do not appear to adversely affect overall QoL, as shown by significant improvement in several disease-related symptoms. There was no evidence to suggest differences in QoL between erlotinib and docetaxel or indeed any of the monotherapy treatment options. We believe that the lack of significant findings with respect to QoL between active treatments in this setting is not surprising and the treatments and their side-effect profiles being compared need to be considered when evaluating the results. For example, in studies comparing monotherapy regimens (TITAN [docetaxel or pemetrexed vs erlotinib] and JMEI [docetaxel vs pemetrexed]), no significant difference between treatments might be an unsurprising and expected result. However, for studies comparing monotherapy to combination therapy, no significant difference in QoL between treatments may be a positive outcome, indicating that the combination treatment is not associated with an additional burden from adverse events that impacts on QoL. Importantly, nintedanib in combination with docetaxel has been shown to improve OS compared with docetaxel alone in patients with advanced adenocarcinoma of the lung after first-line chemotherapy. Comprehensive analysis of the QoL findings from the LUME-Lung 1 trial indicates that prolongation of OS with the addition of nintedanib to docetaxel was achieved without significant impact on patient QoL and, as such, combination treatment with nintedanib plus docetaxel represents an attractive treatment option for patients with NSCLC with adenocarcinoma histology compared with docetaxel monotherapy [4,17]. Of note, the CheckMate 017 study did report a significant improvement in time to first disease-related deterioration on the LCSS 3-item index with nivolumab compared with docetaxel. As this scale was not used in other trials identified in this review, the findings are difficult to interpret with respect to other treatments.

Previous research has shown that the robustness of QoL reporting in general cancer clinical trials has improved over time; the robustness of trials conducted after 2000 compared with those conducted before 2000 was greatly improved [29]. This led the researchers to conclude that, as the robustness of reporting improves, QoL data will increasingly impact on clinical decision-making and treatment policies. Our review of the literature in NSCLC suggests a similar improvement in QoL research, with more recent trials generally providing more comprehensive evaluation of QoL. Limitations of our review should be considered. We did not conduct a formal systematic review of the data and formal assessment of the potential for bias within the included trials was not conducted. It should also be re-iterated that the QoL results from the CheckMate 017 study have not yet been published in full.

Conclusion

It is well recognized that QoL issues are important in advanced NSCLC as it is an incurable disease [30] and this is especially important the more advanced the treatment line. Our review shows that, although QoL methodology varies significantly between studies conducted in this setting, there is a trend towards improved reporting. This is important as full reporting of QoL findings greatly increases the value of the data and improves the reliability of the data for use in clinical decision-making. Consistent use of standard instruments such as the EORTC QLQ-C30 or the FACT-L in future research would not only increase the reliability of findings from a single study, but also allow comparisons across studies more readily.

Although accurate and reliable instruments for QoL measurements are available, studies evaluating QoL in patients with cancer are often criticised regarding their QoL methodologies and reporting of results [25,31]. Criticism may relate to the use of inappropriate instruments, inadequate investigation time and/or inappropriate evaluation methods. When conducting QoL evaluations, it is therefore essential that the instrument being used is suitable for the patient population being examined and the goals of treatment.

Future perspective

QoL is now recognized as an important end point to include in studies evaluating the second-line treatment of advanced NSCLC of adenocarcinoma histology. As shown here, the quality of reporting varies between trials but there is a trend towards improved reporting standards. Given that the second-line treatment of NSCLC is a rapidly evolving area of oncology, especially with the increasing availability of immunotherapies as a treatment option, we believe that QoL will continue to be an important factor in determining treatment choice in this population. As such, we expect that the quality of research in this area will continue to improve, as will reporting standards.