Quality of life endpoints in cancer cachexia clinical trials: Systematic review 3 of the cachexia endpoints series

Abstract The use of patient‐reported outcomes (PROMs) of quality of life (QOL) is common in cachexia trials. Patients' self‐report on health, functioning, wellbeing, and perceptions of care, represent important measures of efficacy. This review describes the frequency, variety, and reporting of QOL endpoints used in cancer cachexia clinical trials. Electronic literature searches were performed in Medline, Embase, and Cochrane (1990–2023). Seven thousand four hundred thirty‐five papers were retained for evaluation. Eligibility criteria included QOL as a study endpoint using validated measures, controlled design, adults (>18 years), ≥40 participants randomized, and intervention exceeding 2 weeks. The Covidence software was used for review procedures and data extractions. Four independent authors screened all records for consensus. Papers were screened by titles and abstracts, prior to full‐text reading. PRISMA guidance for systematic reviews was followed. The protocol was prospectively registered via PROSPERO (CRD42022276710). Fifty papers focused on QOL. Twenty‐four (48%) were double‐blind randomized controlled trials. Sample sizes varied considerably (n = 42 to 469). Thirty‐nine trials (78%) included multiple cancer types. Twenty‐seven trials (54%) featured multimodal interventions with various drugs and dietary supplements, 11 (22%) used nutritional interventions alone and 12 (24%) used a single pharmacological intervention only. The median duration of the interventions was 12 weeks (4–96). The most frequent QOL measure was the EORTC QLQ‐C30 (60%), followed by different FACIT questionnaires (34%). QOL was a primary, secondary, or exploratory endpoint in 15, 31 and 4 trials respectively, being the single primary in six. Statistically significant results on one or more QOL items favouring the intervention group were found in 18 trials. Eleven of these used a complete multidimensional measure. Adjustments for multiple testing when using multicomponent QOL measures were not reported. Nine trials (18%) defined a statistically or clinically significant difference for QOL, five with QOL as a primary outcome, and four with QOL as a secondary outcome. Correlation statistics with other study outcomes were rarely performed. PROMs including QOL are important endpoints in cachexia trials. We recommend using well‐validated QOL measures, including cachexia‐specific items such as weight history, appetite loss, and nutritional intake. Appropriate statistical methods with definitions of clinical significance, adjustment for multiple testing and few co‐primary endpoints are encouraged, as is an understanding of how interventions may relate to changes in QOL endpoints. A strategic and scientific‐based approach to PROM research in cachexia trials is warranted, to improve the research base in this field and avoid the use of QOL as supplementary measures.


Introduction
Cancer cachexia is a multifactorial syndrome defined by an ongoing loss of skeletal muscle mass (with or without loss of fat mass), that cannot be fully reversed by conventional nutritional support and leads to progressive functional impairment. 1 Cachexia in patients with cancer is very common, 2 with a complex pathophysiology and multifaceted impact on patients.To date, there are no universally accepted endpoints for interventional cancer cachexia trials, and endpoints used remain highly variable.Yet if cancer cachexia is optimally treated, then this may have a direct or indirect effect on patients' quality of life (QOL) as studies have shown that improved nutritional status and/or an attenuation of inflammation correspond to improved QOL and well-being, and better mental status. 3,4he terms QOL and health related quality of life (HRQOL) are often used interchangeably 5 as both denote the overall well-being and health aspects in life.These cover broad topics, such as health status, physical functioning, symptoms, psychosocial adjustment, wellbeing, life satisfaction, and happiness, 6 although some claim that HRQOL measures may more appropriately capture changes pertaining to health problems Broadly, both are multidimensional concepts representing an individual's perception of physical, psychological, and social aspects, and overall health (henceforth referred to as 'QOL').These QOL measures fall under the umbrella of patient-reported outcome measures (PROMs) and regulatory agencies (US Food and Drug Administration (FDA), 7 European Medicines Agency (EMA) 8 ) recognizing PROMs as approvable endpoints in evaluating treatment efficacy 9 in other conditions.To date, guidance on specific QOL measures as approvable endpoints in cachexia from regulatory agencies is not clear.
PROMs supplement clinician observations and objective findings with information based on patients' own lived experience.As such, the routine integration of PROMs within clinical research aligns with patient-centred care, defined as 'care that is respectful of, and responsive to, individual patient preferences, needs and values, and ensuring that patient values guide all clinical decisions'. 10PROMs have been utilized throughout cancer clinical trials, as endpoints, interventions, and prognostic markers.For example, PROMs defined the impact of integrating palliative care early in patients with advanced cancer demonstrating improved QOL, psychological distress, symptom burden, 11 and a survival benefit. 12Additionally, empirical evidence indicates that PROMs provide independent prognostic information on survival in several cancer populations. 13Thus, using PROMs within clinical trials in patients with cancer is highly clinically relevant, is well accepted 9 and particularly relevant to patients experiencing the multifaceted impacts of cancer cachexia.
Several types of PROMs exist, for example, the Edmonton Symptom Assessment System (ESAS), 14 the M.D. Anderson Symptom Inventory (MDASI), 15 the Spitzer Uniscale 16 and the early Priestman and Baum LASA scales 17 that all include assessments of wellbeing.Most QOL measures are multidimensional questionnaires, comprising several items that form specific scales, for example, physical, and emotional functioning, supplemented with single items.The questionnaires may measure generic QOL such as the Short Form-36, 18 and the EuroQol-5D (EQ 5D) 19 or may be disease-or condition-specific, with the most frequent cancer-specific PROMs being the Functional Assessment of Cancer Therapy scale (FACT-G), 20 the European Organization for Research and Treatment Quality of Life Questionnaire (EORTC QLQ-C30), 21 and the palliative care EORTC QLQ C15-PAL, 22 the early Rotterdam Symptom Checklist 23 and the Japanese QOL-ACD. 24n terms of what has been used to measure QOL in cachexia trials there are various assessments.The EORTC QLQ-C30 is often supplemented with condition specific measures such as the one for Head-and-Neck Cancer 25 corresponding to the FACIT condition specific measures 26 used together with FACT-G 20 such as the FACT Fatigue and Anemia scales 27 and the FACT Head and Neck Symptom inventory (FHNSI-2). 28The content covered in these validated measures is relevant to patients with cachexia and they are commonly used together with more cachexia specific measures such as the first and subsequently revised Functional Assessment of Anorexia/Cachexia Therapy (FAACT) questionnaire 29 and the EORTC QOL cancer cachexia questionnaire (EORTC QLQ-CAX24). 30Despite these cachexia specific QOL assessments being available, there is no consensus about the most appropriate QOL endpoint in cachexia trials with inconsistency of assessments being used, analysis measures differing and subsequently varying reporting approaches.There is also no robust evidence to support which might be easiest to use in a trial and/or preferred by trial participants.These limitations are further compounded by the lack of a widely accepted 'minimally clinically important difference (MCID)', and this then impedes trial design and ultimately drug development.
This systematic review is part of a series of reviews assessing endpoints in cachexia clinical trials and aims specifically to examine QOL.The main objective was to describe the frequency and variety of QOL endpoints.This review includes descriptions of trial characteristics, interventions, QOL measures, reporting of QOL, and the relationship with significant primary and/or secondary outcomes.

Protocol and registration
This systematic review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyzes Statement PRISMA (Supporting file S1). 31

Search strategy
The search for trials published from January 1990 until 2 June 2021, was conducted by a research librarian (University of Oslo, NO) for this review series in the following databases MEDLINE (Ovid), EMBASE (Ovid) and the Cochrane Central Register of Controlled Trials.The search was registered on the International Prospective Register of Systematic Reviews (PROSPERO -CRD42022276710) where further details are available. 32The full electronic search strategy including limits used for the OVID Medline database can be found as Supporting file S2.
The systematic review is part of a comprehensive collaboration including six reviews examining different endpoints in cachexia (body composition, oncology, physical function, PROMs, systemic inflammation, and nutritional).One search was performed for all reviews followed by central appraisal, data extraction and quality assessment.Thereon, eligible trials were reviewed and those specifically examining quality of life were included in the present review.For the present review, the search was updated from 2 June 2021 to 17 October 2023.

Eligibility criteria
Articles were considered eligible if they were controlled trials investigating interventions which aimed to treat or attenuate cachexia (defined as detailed in PROSPERO) in adults with cancer.There were no restrictions on the type of intervention (pharmacological, nutritional, exercise, multimodal, etc.) or type of comparator.To reduce bias and focus on outcomes with the most clinical impact, trials were excluded if they had randomized fewer than 40 patients, and the intervention lasted <14 days.
For the present review on QOL, some additional inclusion criteria were applied: • Patient reported QOL (used interchangeably with HRQOL) should be a stated outcome • Use of validated QOL measures, not ad-hoc measures • Studies where QOL partial domains of PROMS (e.g., EORTC emotional functioning) were used were eligible The following exclusion criteria were applied: • Insufficient reporting of QOL (i.e., data not shown, not compared between intervention and control groups, or lack of appropriate statistical measures) • Trials using observer-rated measures of physical functioning, for example the Karnofsky Performance Status scale (KPS), 33 or Eastern Cooperative Oncology Group Performance Status scale (ECOG) 34 as a substitute for self-reported QOL • The use of a single symptom scale denoting (e.g., assessing appetite and fatigue) conceptualized as a measure of QOL

Data selection and extraction
All articles identified were transferred to Covidence software. 35Article selection based on titles and abstracts was completed by three researchers in the core team (B.L., T. S .S., and O. F. D.).Any uncertainties in assessing the eligibility of the trials were discussed among the authors until a consensus was reached.A data extraction table was developed, pilot-tested and refined within the review group before data were extracted from each article by two independent authors from the review group.Articles relevant to each systematic review were then identified from the data.For this paper, relevant articles assessed the specified QOL endpoints noted in this review.

Assessing the risk of bias
The methodological quality of each study was systematically assessed by four independent reviewers (J.M. D., J. S., B. L., and O. F. D.) with the Modified Downs & Black Scale. 36he measure assesses among other criteria, study design, blinding, sample size, estimate of variance reporting, and whether the outcome is defined and robust.

Outcomes
This systematic review examines the assessment of QOL in RCTs using validated PROMs on QOL as study endpoints.
More specifically, it describes the following: • the number of identified cancer cachexia RCTs stating QOL as a primary or secondary, or identified as an exploratory outcome;

• study characteristics and interventions;
• the QOL measures used, including content and properties related to validation, international applicability, mode of administration, dimensionality, scoring methods and interpretation; • the reporting of QOL results, including statistical methods; and • correlations with significant primary or secondary study outcomes, as appropriate.

Data analyses
As expected, the number of retrieved trials was large and heterogeneous.Given this volume and with the main objective being to describe the frequency and diversity of QOL endpoints used, a meta-analysis of the effect of the interventions was not relevant.Hence, the data were summarized narratively.In trials reporting significant findings on any QOL parameter, raw scores on these subjective measures and the corresponding variability were extracted (if available) to enhance the interpretation of results.

Identified cancer cachexia clinical trials with quality of life as an outcome
The systematic literature search for the series of reviews on cachexia outcomes identified 8166 trials (Figure 1).After deleting duplicates, 7435 papers were retained to screen abstracts, producing 387 articles for full-text reading.

Study characteristics
The characteristics of the 50 included trials with QOL outcomes are reported in Table 1.These were published between 1996 and 2023, and conducted in 20 different countries, most often the United States and China (both = 6) followed by Italy, Australia, and Iran (n = 4 for all).Three trials were multinational. 38,50,78The total sample size based on the number of randomized patients was 6893, but varied considerably across trials, ranging from 42 42 to 469. 56st trials (39/50, 78%) included multiple diagnostic groups.Thirty-one trials mentioned all cancer diagnoses involved, while eight were less specific using broad terms such as gastrointestinal or advanced cancer (Table 1).Twenty-five of the 50 trials (50%) included patients with lung cancer while pancreatic cancer (42%) was the second most common diagnostic group; Lung (n = 6) and pancreatic cancers (n = 4) were the two diagnoses most used in the trials limited to one cancer type (Table 1).

The interventions Pharmacological interventions dominated (27/50, 54%
) with diverse pharmacological agents, that is, anticancer drugs, appetite stimulants, anti-inflammatory drugs, and dietary supplements (Table 1).Seventeen trials (34%) were categorized as nutritional interventions, and composed with different nutritional agents, and dietary counselling.Nutritional interventions were, for example, whole-course nutritional management programme provided by a specialized or multiprofessional teams, 45,71 protein and energy-dense oral nutritional supplement with n-3 fatty acids, 38,75 whey protein isolate supplements, 44 or thorough follow up of nutritional status with tube feeding or parenteral nutrition as necessary. 74,81Nutritional advice was also included in some of the six multimodal programmes, for example, the cognitive behavioural intervention by Britton et al. 41 and combined with physical exercise. 77,80The median duration of the interventions was 12 weeks (range 4-96).

Study outcomes
,82,83,86 QOL was the single primary outcome in six of these trials 39,40,47,55,80,83 and one of three or four co-primary outcomes in the remaining nine (Table 1).Five of the 15 trials (27%) defined a clinically meaningful change for QOL, either as a 5 or 10% change in the scales or item scores 38,40,80 or specified as a difference of .45 or .5 SD. 39,83 All except four 37,39,40,55 of these 15 trials used the EORTC QLQ-C30, either alone (n = 7) or in combination (n = 4) with other PROMs (Table 1).
The reporting of QOL results varied.Mean (standard [SD]) scores with corresponding p-values for patient groups were used in eight trials. 47,54,62,64,78,80,83,86Two trials reported mean (standard error of the mean [SEM]) or mean (95% confidence interval [CI]) values 38,54 and one presented the median and range of scores. 69Four trials 39,40,55,82 reported the absolute or per cent change in mean scores at the different assessment points, while one study presented both mean (SD) and per cent change. 391][72]76,77,79,84,85 Four of these specified a clinically significant difference for the QOL measures, being a 10%, 20% or 25% difference on the 0-100 scales between groups.This difference was assessed either at a specific assessment point or as a within-group change over time. 49,61,72,74ometimes QOL measures were not specified as a study objective even if the QOL results were presented in the re-sults section.However, if the latter applied these data were assessed and we defined QOL as an exploratory outcome in four trials. 42,51,57,73None of these trials defined a clinically significant difference for the QOL measures.

The quality of life measures
Seventeen different QOL measures were used in these 50 trials.Two of these, the SF-36 18 and the EQ-5D 19 are generic QOL measures while the remaining 15 are cancer-specific.The most commonly used measure was the EORTC QLQ-C30 21 in 60% of the trials, while 17 trials (34%) used different versions of the FACIT.Thirteen trials (26%) used multiple measures of QOL, often including diagnosis or condition specific measures, such as the EORTC H&N35 25 and the anaemia and fatigue FACIT measures. 27FAACT was the only cachexia-specific measure used, either in the 18-or 12-Item versions. 29Supporting file S3 presents the measures used, their content, assessment period, scoring, number of scales and items and whether a summary measure could be calculated.Figure 2 indicates how often different measures were reported together.With the exception of the measure developed in Japan by Kurihara et al., 24 all measures were validated in an international context and demonstrated cross-cultural applicability.

Trials reporting statistically significant quality of life results
Eighteen trials reported statistically significant QOL benefits in favour of the intervention arm. 37,39,41,45,49,52,54,56,60,62,64,66-68,74,75,79,82 Nine of these 18 studies (50%) used pharmacological interventions, and had a total sample size of 2895 (ranging from 47 to 469).The EORTC QLQ-C30 was the most common measure; used in 61% (11/18) of the trials.The length of the intervention in these trials varied from eight to 28 weeks.Two trials had a pre-set definition of a clinically significant difference, that is, a difference of 10 points or more on the EORTC-QLQ-C30 74,52 and on the QLQ-H&N35. 74L was the primary outcome in six (32%) of these trials 37,39,54,62,64,82 and a secondary outcome in 12. 41,45,49,52,56,60,[66][67][68]74,75,79 Authors' interpretation of the QOL results are summarized in Table 2, with the statistical presentation of significant results in Table 3.None of these 18 trials reported statistical correlations between the QOL outcomes and other outcome measures.If a potential relationship was mentioned, this appeared in the discussion section and was vaguely described as 'being associated with' symptom items or other from the intervention endpoints, for example weight gain in the questionnaires.
Only three of the 18 trials presenting significant results had defined a magnitude of a statistically and/or clinically significant difference, that is, a 0.45 SD corresponding to an 11% change on the 0-100 overall LASA or Uniscale scores 39 or a difference of 10 points or more on the EORTC-QLQ-C30 measures. 52,74None of the trials reported adjustments for multiple testing in the statistical significance analyses, even if most QOL measures were composed of several items and domains.

Discussion
This review identified 50 RCTs in cancer cachexia where QOL was assessed as an outcome.Overall, 18 trials reported  1.  statistically significant differences in QOL outcomes, in favour of the intervention groups.Of these, six had QOL as the primary study outcome, and 12 had it as a secondary outcome.These findings, although encouraging, indicate many considerations are needed when incorporating QOL in cachexia clinical trials.

Studies
Firstly, defining what a clinically significant difference represents is challenging, and this was seldom reported.Only one trial (QOL was the primary outcome) defined a clinically meaningful difference in QOL (11%/0.5 SD), 39 while another used a 10% difference on the 0-100 numerical scales, but did not specify which of the multiple outcomes this applied to. 82A 'rule of thumb' is that a difference of 7-15% on the 0-100 scale, or a 0.5 SD is meaningful to patients. 87,88owever, the difference between minimally clinically important differences (MCID) at a patient level versus at the group level is not clear.The latter relates to mean differences between groups or mean change over time reaching a level of significant difference, whereas individual patient change over time categorize, for example, non-responders/responders to a particular treatment effect is the focus at the patient level.These approaches require different thresholds for correct interpretations as emphasized in ongoing international projects aiming to standardize the measurement and interpretation of PROMs. 89,90As cachexia is a multifactorial syndrome, it is important to understand how changes in QOL relate to changes in other endpoints.For example, does improved QOL correlate with improved physical function and vice versa?An understanding of such relationships is critical both for patient benefit and also to know how pathophysiological changes (and therefore potential mechanisms of action of interventions) relate to changes in endpoint(s).None of the 18 studies where QOL improved examined how this related to other endpoints.This represents an area that should be addressed in future trials.
Secondly, sample size calculations need to be applied when QOL endpoints are assessed, although QOL improved in a proportion of studies, sample size estimations in relation to this was uncommon, as were effect sizes.
Thirdly, the optimal time point for measuring QOL needs to be clarified.Usually, these are assessed over time with multiple QOL assessments, and this was the case for some trials included where for example QOL improved after 12 weeks.This finding could mean that a significantly improved QOL at 4 weeks may be sustainable for the next 2 months as well, that it was a random finding, or maybe that it was not attributable to the intervention per se but to other factors influencing QOL.Yet other factors may impact QOL and as does the expected deterioration in patients with cancer cachexia. 3inally, the complex intervention(s) complicates the interpretation of results as disentangling which affects which outcome, is challenging; particularly with QOL.Yet these multimodal interventions in cachexia trials are recommended in   cachexia treatment 91 and practical guidelines. 3Additionally regulatory bodies advocate QOL endpoints, so ways to integrate appropriately and assess QOL in cachexia trials is essential and a research priority.Taken together, much work remains to integrate QOL measures optimally and meaningfully into cancer cachexia clinical trials.Some proposals can be found in Supporting file S4.
As QOL assessment is likely to remain a central tenet of the cachexia trial endpoint spectrum, the question remains as to which measure should be used.We noted that the EORTC QLQ-C30 21 was the most frequently used measure (as evidenced in other reviews 4,92 followed by the FACT-G 20 (part of the specific FACT-modules).These are multidimensional, internationally validated and developed through rigorous and stepwise scientific processes.The EORTC QLQ-C30 and FACT-G have been adapted to include cachexia-specific QOL assessment via the EORTC Cachexia-24 module 30 and the 12-item FACIT cachexia-specific instrument (FAACT). 29Whilst these adaptions are welcome further evaluation of these is needed before they can be recommended as being the preferred QOL assessment; specifically, regarding response to change in patients with cachexia. 4It could also be proposed that single items from QOL assessments could be assessed or even single items from multiple assessments combined; yet one of the limitations of such an approach is that when tools are dissected in terms of their component parts, the validity is often questioned, and these tools have usually been developed and assessed as a whole.
It is acknowledged that patients with cancer cachexia are often frail or deteriorate rapidly.Thus, the balance between the need for short measures to reduce patient burden and the need for comprehensive assessments may be challenging.However, technological development with digital measures and computer adaptive testing methods of the EORTC and FACIT measurement systems that tailor the questions to the individual patients, represents a major step forward in the monitoring and follow-up of patients, in research and clinical practice.
Both the EORTC QLQ-C30 and the FACT-G/FAACT have composite scores, calculated as the mean of the combined scale and item scores for the EORTC QLQ-C30, and by totalling subscale FACT-G scores.For the EORTC QLQ-C30; however, the distinction between the Global QOL score and the composite score is important, as the former consists of only two items that combine physical health and the patient perception of overall QOL.Only one reviewed paper used the composite EORTC score, while the Global QOL Score was used as the only outcome measure in two of the nine trials reporting significant QOL results with QOL as a primary outcome 62,82 and supplemented with the physical functioning score. 54This is problematic because the Global QOL scale score may not correspond well to specific item or scale scores, as it does not appear to be sufficiently sensitive.A probable explanation might be a response shift over time in patients with advanced disease: 'taken together my situation is not that bad', despite reporting a relatively high symptom burden.On the other hand, it is acknowledged that an improvement in one specific scale within a multidimensional QOL measure denotes a generally improved QOL.The well-validated tools contain both single items and multidimensional scales assessing different aspects of QOL including symptom burden, and possess a reasonable sensitivity and specificity.Thus, a sole focus on symptoms such as appetite or fatigue should not be used alone to indicate a multidimensional construct such as QOL.Thus, trials that used a single item to denote QOL in this review were not included.Also, it is discouraging that associations between QOL outcomes and other more objective results were not emphasized in any of the reviewed trials.Further, trials using weight loss or gain as outcomes face several challenges that were rarely elaborated on in the trials.Examples are the variability in measures (per cent, kilograms, and slopes on the weight curves) and the association with body composition variables that may affect muscle mass, strength, and functioning that may affect specific QOL scores, to which a global score is not sufficiently sensitive.
Removing confounding factors, particularly in the context of a clinical trial where QOL is measured, is worthy of note.Of the 15 trials reporting significant QOL differences across groups, nine were pharmaceutical trials, while the remaining six had a more individualized approach.It can be hypothesized that a direct patient-centred approach with individual or group follow-ups in terms of meetings, exercise groups, frequent phone calls, or digital consultations may improve the patients' emotional wellbeing and mental state.However, this 'attention' effect would be controlled for, though not blinded against, if there were some sort of active intervention in the control group, as opposed to conventional care.To our knowledge, the direct benefits of being in a study are rarely investigated, but it is likely that being 'seen' is a positive factor.In cachexia trials where counselling on nutrition and physical activity are commonplace, therapeutic relationships will develop and these are likely to influence QOL for trial participants.It is also reasonable to assume that other aspects may influence QOL, independent of the intervention being assessed.Examples include psychological distress caused by disease progression, side-effects from anti-cancer therapy, or financial worries.Disentangling these facets from a 'pure' QOL assessment in a clinical trial is complex and may be challenging to truly understand.

Strengths and limitations
A key limitation of the review was that the heterogeneity of trial designs, populations studied, and variation in interventions prevented direct comparisons of results or a meta-analysis.This limitation was partly due to the decision to include trials that assessed QOL in varying hierarchies of endpoints.As such, where QOL was not a primary endpoint, the trial was not powered to conclude on QOL results.This approach was justified to ensure that no important information was missed.
The multifactorial origin of cancer cachexia calls for multimodal interventions, even if it makes it difficult to prove an exact relationship between interventions and outcomes.Related to this is that the use of aggregated data limits a detailed assessment of the relationship between different QOL measures and other study outcomes.It is therefore not possible to draw firm conclusions regarding which measure to use, given the QOL measures are multidimensional and their sensitivity to changes in other outcomes is unknown.We also have to acknowledge that in patients with cachexia, other symptoms related to cancer will be common, in addition to co morbidities, in older patients in particular.Choice and use of QOL instruments within the context of clinical trials must be cognizant of these factors.
We purposefully chose to use a quality assessment tool that was general rather than specific to certain endpoints, and we regard this as an appropriate methodology.Further, we decided to do quality assessments at the initial level by reviewers to limit bias.We believe that the selection of another approach for reporting quality assessment endpoints would have had only a minor influence on our conclusion.
It was challenging to distinguish between studies trying to prevent or treat cachexia per se (involuntary weight loss) versus those trying to treat the symptoms caused by cachexia.There was no uniform definition of cachexia used throughout to the extent that some studies examined cancer anorexia (a component of cachexia syndrome) versus targeting lean mass versus targeting multiple components.Indeed, the lack of uniform trial definition reflects the current status quo of cachexia clinical practice whereby there are several, alternative operational diagnostic criteria (e.g., Fearon definition) 1 or the Global Leadership Initiative in Malnutrition (GLIM) criteria 93 each of which has been established by expert consensus alone.From a QOL perspective however, it should be acknowledged that objective changes such as weight gain or improved performance status might contribute to improvement in one or more QOL aspects, even if it does not change the patient's cachectic state.Future work should be clear as to the primary aim of any intervention as potentially the term 'cancer cachexia' may be too vague in the context of a clinical trial intervention.
One study strength is the review of trials using PROMs that span more than 30 years of research, coupled with the fact that the most frequently used and well-validated QOL measures in the reviewed trials date back to the early 1990s, that is, the EORTC QLQ-C30 and FACT-G.The EORTC QLQ-C30 was used in the two oldest trials in this review, published in 1999.Also, the thorough approaches used for evaluating scientific quality, extraction, and appraisal of papers is a major strength.This also applies to the careful registration of relevant variables in a common database for a series of reviews of cachexia trials involving double or triple appraisals for paper retention.

Conclusions
QOL is an important endpoint in cancer cachexia trials, regardless of whether an improvement is due to direct effects from a specific drug or results from synergistic effects of the multiple components in complex interventions.Thus, it makes sense to include patient-reported QOL endpoints in cancer clinical trials.As demonstrated in this review, however, comprehensive descriptions of the patient samples and characteristics varied, as did the presentation of statistical considerations related to sample size, power estimations, presentation of results and correlations with other study outcomes, and adjusting for multiple significance testing.
Thus, we call for a more rigorous approach to assessing QOL as an endpoint in cancer cachexia trials, including defining what a MCID is, how QOL relates to mechanism of action of the intervention, other key endpoints (e.g., physical function), and learning from other areas where regulatory approval has been given on the basis of a PROM of QOL.As cancer cachexia has a profound impact on patients' QOL, and as it is a multidimensional construct, we recommend the use of well-validated comprehensive QOL measures with cachexia-specific modifiers and advise against using single items as surrogate indices of QOL.Taken together, these will inform future trials and clinical practice.

Figure 1 a
Figure 1 PRISMA flow chart.

Figure 2
Figure 2 Network diagram reporting of QOL measures.The size of the circles represents the frequency of each measure, and the width of the connecting lines reflects the number of studies reporting each pair of measures.The measures QOL-ACD, RSCL, and EORTC QLQ-C15-PAL are not shown as these have not been presented in combination with other measures of QOL.**FAACT includes both the 12-and 18-item version of this measure.Numerical details are reported inTable 1.
with QOL as their primary outcome (N a statistically significant difference across groups or within groups over time.Significance is in favour of the intervention group, unless otherwise stated.2c Statistics used for QOL scores.13 CF, cognitive functioning (function scale EORTC QLQ-C30); GEE, generalized estimating equation; PG-SGA, patient-generated subjective global assessment; PF, physical functioning (function scale EORTC QLQ-C30); RF, role functioning (function scale EORTC QLQ-C30); RT, radiation therapy; SD, standard deviation; SF, social functioning (function scale EORTC QLQ-C30).

Table 2 (
a Number of randomized patients.b Defined as a statistically significant difference across groups or within groups over time.Significance is in favour of the intervention group, unless otherwise stated.c Statistics used for QOL scores.CF, cognitive functioning (function scale EORTC QLQ-C30); GEE, generalized estimating equation; PG-SGA, patient-generated subjective global assessment; PF, physical functioning (function scale EORTC QLQ-C30); RF, role functioning (function scale EORTC QLQ-C30); RT, radiation therapy; SD, standard deviation; SF, social functioning (function scale EORTC QLQ-C30).

Table 3
Statistical presentation of significant QOL resultsDefined as a statistically significant difference across groups or within groups over time.Significance is in favour of the intervention group, unless otherwise stated.
b c Statistics used for QOL scores.