Biomarker endpoints in cancer cachexia clinical trials: Systematic Review 5 of the cachexia endpoint series

Abstract Regulatory agencies require evidence that endpoints correlate with clinical benefit before they can be used to approve drugs. Biomarkers are often considered surrogate endpoints. In cancer cachexia trials, the measurement of biomarkers features frequently. The aim of this systematic review was to assess the frequency and diversity of biomarker endpoints in cancer cachexia trials. A comprehensive electronic literature search of MEDLINE, Embase and Cochrane (1990–2023) was completed. Eligible trials met the following criteria: adults (≥18 years), prospective design, more than 40 participants, use of a cachexia intervention for more than 14 days and use of a biomarker(s) as an endpoint. Biomarkers were defined as any objective measure that was assayed from a body fluid, including scoring systems based on these assays. Routine haematology and biochemistry to monitor intervention toxicity were not considered. Data extraction was performed using Covidence, and reporting followed PRISMA guidance (PROSPERO: CRD42022276710). A total of 5975 studies were assessed, of which 52 trials (total participants = 6522) included biomarkers as endpoints. Most studies (n = 29, 55.7%) included a variety of cancer types. Pharmacological interventions (n = 27, 51.9%) were most evaluated, followed by nutritional interventions (n = 20, 38.4%). Ninety‐nine different biomarkers were used across the trials, and of these, 96 were assayed from blood. Albumin (n = 29, 55.8%) was assessed most often, followed by C‐reactive protein (n = 22, 42.3%), interleukin‐6 (n = 16, 30.8%) and tumour necrosis factor‐α (n = 14, 26.9%), the latter being the only biomarker that was used to guide sample size calculations. Biomarkers were explicitly listed as a primary outcome in six trials. In total, 12 biomarkers (12.1% of 99) were used in six trials or more. Insulin‐like growth factor binding protein 3 (IGFBP‐3) and insulin‐like growth factor 1 (IGF‐1) levels both increased significantly in all three trials in which they were both used. This corresponded with a primary outcome, lean body mass, and was related to the pharmacological mechanism. Biomarkers were predominately used as exploratory rather than primary endpoints. The most commonly used biomarker, albumin, was limited by its lack of responsiveness to nutritional intervention. For a biomarker to be responsive to change, it must be related to the mechanism of action of the intervention and/or the underlying cachexia process that is modified by the intervention, as seen with IGFBP‐3, IGF‐1 and anamorelin. To reach regulatory approval as an endpoint, the relationship between the biomarker and clinical benefit must be clarified.


Introduction
][3][4] Treatment options remain limited, and this is in part due to sub-optimal clinical trial design, including a lack of clarity on the optimal endpoints to use.Endpoints in cancer cachexia trials can be split into various categories, 5 often aligned with the various cachexia definitions. 3Yet using different endpoints simultaneously in trials can lead to conflicting results.To illustrate, the ROMANA trials 6 found that anamorelin increased lean body mass but did not improve hand grip strength.There are multiple reasons why this may have happened, including endpoint sensitivity, the mechanism of action of the trial drug, temporal aspects or perhaps mainly population selection.A post hoc pooled analysis of these trials showed that participants who had systemic inflammation (measured using C-reactive protein [CRP] and albumin, combined in the modified Glasgow prognostic score [mGPS]) had a better response to anamorelin, both considering lean body mass and hand grip strength, than those who did not. 7Recently, trials examining ponsegromab, 8 which targets growth differentiation factor 15 (GDF-15), have used elevated levels of GDF-15 as both an entry criteria and an exploratory endpoint.These two examples provide some rationale for the role of biomarkers in cancer cachexia clinical trials; however, these are the exception rather than the rule.Further work in this area is required to help facilitate future trials.
The US Food and Drug Administration (FDA) states that endpoints can be either a clinical outcome, known as a direct measure (such as overall survival), or, in recent years, a surrogate endpoint. 9A biomarker is an example of a surrogate endpoint and is defined as an objective indication of the current medical state of an organism. 10For a surrogate endpoint to lead to the approval of a new drug, the endpoint must have extensive evidence to support its correlation with clinical benefit, 9 and the biomarker must be reliably assessable, predictive and responsive. 11The FDA has approved several biomarkers to study the outcome of pharmacological interventions, such as serum insulin-like growth factor 1 (IGF-1) levels in patients with acromegaly. 12However, to date, no biomarker assayed from blood or body fluid has been featured as the primary endpoint for cancer drug approval, but they have formed part of a composite endpoint. 13In cancer cachexia, the picture is still less clear as an appraisal of endpoints has not been undertaken.Furthermore, it could be argued that there is less of a published, demonstrable relationship between cachexia biomarkers and clinical benefit.
The aim of this systematic review is to explore which biomarkers have been used in cancer cachexia trials and with what frequency and diversity.

Methods
This systematic review is reported in accordance with the Preferred Reporting for Systematic Reviews and Meta-Analyses (PRISMA) statement. 14The review protocol was prospectively registered at the International Prospective Register of Systematic Reviews: PROSPERO (CRD42022276710). 15This systematic review is part of a collaboration reviewing different endpoints in cancer cachexia trials. 16,17arch strategy A systematic search of the MEDLINE (OVID), Embase (OVID) and Cochrane Central Register of Controlled Trials databases was conducted by a senior research librarian (University of Oslo), and studies from 1 January 1990 to 17 October 2023 were assessed.Search results were synthesized and managed using the web-based systematic review software 'Covidence' (Veritas Health Innovations, Melbourne, Australia), and duplicates were removed.A detailed search strategy is outlined in Appendix S1. not included.All routine haematology and biochemistry tests that were explicitly stated as being used to measure intervention toxicity/safety were not included.In addition, routine haematology and biochemistry tests stated to have been obtained from the participant in the study methodology but not reported were also not included.Biomarkers used to measure the compliance of an intervention were considered separately.Inclusion was irrespective of the site of primary malignancy, modality of intervention (e.g., pharmacological, nutritional and physical exercise) or choice of comparator.Trials were excluded if they studied fewer than 40 patients and/or if the intervention lasted <14 days.All included full-text articles were written in the English language.

Data selection and extraction
The titles and abstracts of the identified studies were independently reviewed by three authors (O.D., T. S. S. and B. L.).Those selected were subsequently subject to full text review (R. D., M. S., M. Y. and J. T.).A pre-defined data extraction table was developed (R. D., M. S. and M. Y.) and pilot-tested before relevant data points were extracted (M.Y.).

Assessing risk of bias
The methodological quality of each study was independently assessed by four reviewers (J.S., J. M., O. D. and B. L.) using the modified Downs and Black checklist. 18This tool assesses several criteria, including study design, internal and external validity and responding standards.A total score of 28 is possible for randomized trials and 25 for non-randomized studies.Previous investigators 19 have classified scores as excellent (26-28), good (20-25), fair (15-19) and poor (<15).

Data analysis
In assessing the frequency and diversity of biomarkers used in cachexia intervention trials, study characteristics, participant details and disease demographics are reported descriptively.Additionally, the large number and heterogeneous nature of the trials included meant that a meta-analysis of treatment effects on each endpoint was not feasible.Visualizations were conducted using RStudio Version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria), with packages including tidyverse.

Results
Figure 1 details the PRISMA diagram.After the removal of duplicates, 7435 records were reviewed by title or abstract (the abstract was assessed where the title was insufficient), resulting in 387 records being appraised in full.Following appraisal, 285 records were further excluded, leaving 129 that were eligible for the systematic review database.Of these, 52 studies analysed biomarker endpoints and thus were eligible to be included (Figure 1).
In total, 99 different biomarkers were used across the 52 different trials (see Appendix S2).Ninety-seven of these biomarkers were assayed from blood.Two (nitrogen and creatinine) were assayed from urine in the same trial, dated 1998. 62Biomarkers were explicitly listed as a primary outcome in six trials (11.5%).Two trials 26,53 used a biomarker as part of their entry criteria, and one trial was powered for a biomarker, tumour necrosis factor-α (TNF-α). 51Only one tumour-specific biomarker, prostate-specific antigen (PSA), was measured across the included trials.It was used in two trials 26,47 and statistically did not change significantly in either.Overall, most biomarkers (n = 53, 53.5%) were used only in one clinical trial.
Figure 2 summarizes the temporal trends in the most commonly studied biomarkers in cancer cachexia trials.Albumin has been used as a biomarker in cancer cachexia trials since at least 1993 and is still used today.As can be seen, readably obtainable biomarkers such as albumin, pre-albumin, platelets, creatinine and haemoglobin have been in use since the 1990s.The use of more specific biomarkers, such as IGF-1, has increased recently.
Figure 3 summarizes the 12 most commonly assessed biomarkers (used in six or more trials) and the most noteworthy biomarker, the Glasgow prognostic score (GPS).In total, 50 trials (96.2%) featured at least one of these 12 biomarkers.Albumin (n = 29, 55.8%; 3512 participants) was the most used biomarker in cancer cachexia trials.Nine of the 29 (31%) trials investigating albumin as a biomarker demonstrated statistically significant changes in serum levels between intervention arms.In five of these trials, 34,35,39,60,62 albumin increased in the intervention arms, presumably as a result of less inflammation and/or improved nutritional status.In three studies, albumin decreased in both the intervention and control arms, but to a lesser extent in the intervention arms of two trials 20,28 and the control arm of another 29 and one study.In the last of the nine studies with statistically significant changes reported in albumin, the data or direction were not presented. 57Six of these trials studied a nutritional intervention (n = 16, 37.5%), two studied a pharmacological intervention (n = 10, 20%) and one studied a multimodal intervention (n = 3, 66.6%).
Another biomarker of note was insulin-like growth factor binding protein 3 (IGFBP-3), which statistically significantly increased in all three trials in which it was studied.These three   QoL, quality of life; RBC, red blood cell count; RCT, randomized controlled trial; RDW, red cell distribution width; SCLC, small cell lung cancer; TIBC, total iron binding capacity; TNF, tumour necrosis factor; VLDL, very low-density lipoprotein; WBC, white blood cell count; ZAG, zinc-α2-glycoprotein.

Table 1 Key characteristics of the included trials
a Listed as a primary outcome or used in the power calculation implying a primary outcome.trials all studied anamorelin, a ghrelin analogue.The same three trials also observed statistically significant increases in IGF-1.Nine other biomarkers had a significant change the single trial in which they were featured.Six of these (docosapentaenoic acid [DPA], alpha lipoic acid [ALA], dihomo-gamma-linolenic acid [DGLA], gamma-linolenic acid [GLA], linoleic acid [LA] and vitamin D) were studied across two trials, 24,46 which both featured a nutritional component to their intervention.Pro-inflammatory cytokines comprised 6 of the 99 biomarkers (6.1%).TNF-α statistically significantly changed in the greatest number of studies (3/14, 12.5% [decreased in two trials 50,52 and increased in one trial 26 but to a lesser extent than control]), followed by decreases in IL-6 levels (2/ 16, n = 12.5%).As seen in Figure 3, TNF-α only changed significantly in studies with a pharmacological intervention, and this was the same for IL-6.A low percentage of statistically significant change was also seen in the leucocyte count (decreased in 1/11 [9.1%]).IL-8 (n = 2), IL-15 (n = 1), IL-1α (n = 1) and IL-1β (n = 1) did not change significantly in any of the trials they featured in.
Three different biomarkers/biomarker scores were calculated from markers of inflammation: the GPS (n = 5), the neutrophil-to-lymphocyte ratio (n = 2) and the platelet-to-lymphocyte ratio (n = 1).Only GPS changed significantly between groups, and this was a decrease in one trial. 52All trials that included GPS used a pharmacological intervention.

Discussion
This is the first systematic review to examine the use of biomarkers as endpoints in cancer cachexia.Many different biomarkers (n = 99) have been used across 52 trials, employing three different interventions (pharmacological, nutritional and multimodal) in at least seven different tumour types.The heterogeneity in trial design is evident, and biomarkers were used as a primary endpoint in six trials, with the remaining being an exploratory endpoint.Only one trial used a biomarker to guide power calculations, and 53.1% (n = 52) of biomarkers were only ever studied once.However, the landscape is shifting, and there is a trend towards the use of more specific biomarkers such as IGF-1 and IL-6.It is not possible to draw any robust conclusions about the many biomarkers presented in this review, as they are not featured frequently enough for patterns to emerge.A comment can be made on the more frequently assessed biomarkers, and in interpreting these findings, statistical significance or lack thereof does not necessarily mean that the biomarker is not useful.However, the effectiveness of the intervention, sample size calculations and sensitivity of the biomarker are all influencing factors.
Hypoalbuminaemia has long been recognized as a feature of cancer cachexia, 72 and it is therefore expected that albumin was the most frequently used biomarker (29/52 trials), likely due to its dual role as a marker of inflammation and nutrition.Significant changes were only seen between trial arms in one third of the studies examining albumin.Due to the long half-life of albumin (approximately 3 weeks), daily protein intake will have little immediate impact on serum levels, 73 and consequentially, nutritional intervention only yielded 3 significant results of the 16 studies in which albumin was assessed following a nutritional intervention.
In pharmacological interventions, changes in albumin seemed to be dependent on the intervention.Studies that were positive used an anti-inflammatory pharmacological intervention: indomethacin and thalidomide, which act to decrease levels of IL-6 74 -albumin synthesis is inhibited by IL-6, 75 and this inflammatory mechanism is related to quality of life. 76These two studies suggest that when albumin is used as a biomarker in trials in which the inflammatory genesis of cachexia is targeted via an anti-inflammatory intervention, it may prove valuable as an endpoint.In contrast, appetite stimulants increased oral intake but did not increase albumin.
With other biomarkers of the systemic inflammatory response, no substantial candidates were remarkable.Twenty-two studies measured CRP as an endpoint, an acute-phase protein with well-documented prognostic value in cancer.Of note, in the two trials with statistically significant CRP results, both interventions included an anti-inflammatory component.One studied indomethacin, 62 and the other used a combination of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), 60 which has been shown to have anti-inflammatory properties. 77However, six other trials utilized a nonsteroidal anti-inflammatory drug (NSAID) (pharmacological = four and multimodal = two) but did not show significant results.This is in keeping with a meta-analysis of randomized controlled trials (RCTs) in patients with rheumatoid arthritis, which showed that NSAIDs have no effect on CRP levels. 78One of the pharmacological interventions used a drug that is not known to have any anti-inflammatory properties (mirtazapine), suggesting that there is a disconnect between the biomarker chosen to be studied, the intervention and the cancer cachexia process.However, recently, the Global Leadership Initiative on Malnutrition consensus group has proposed that confirmation of inflammation should be guided by clinical judgement based upon the underlying diagnosis or condition, clinical signs or CRP. 79Furthermore, given that CRP and its derivatives, including albumin (mGPS and CRP-to-albumin ratio), are obligatory measurements in randomized clinical cancer trials, 80 used extensively in immunotherapy trials, 81,82 and nutritional support will be increasingly given in this context, it is likely that CRP and its derivatives will become an increasingly important measurement in randomized cancer cachexia clinical trials.
When considering pro-inflammatory cytokines, relatively few significant results were seen.A recent systematic review 83 looked at the relationship between cytokines and symptoms in advanced cancer and found correlation.This would suggest that trials need to focus more on measuring biomarkers involved in both the cachexia process and the intervention they have chosen to use.Other challenges with these include sampling errors, detection levels, paracrine/autocrine effects and sample timing.
This disconnect is less evident with regard to IGFBP-3, IGF-1 and anamorelin.Anamorelin is a ghrelin mimetic and growth hormone secretagogue that has been shown to  increase levels of both IGFBP-3 and IGF-1. 84In all three studies, both IGFBP-3 and IGF-1 levels increased significantly in the intervention group.However, as mentioned previously, anamorelin has been shown to improve weight but not hand grip strength, and it is perhaps function that is arguably going to make the biggest improvement to the lives of people with cancer cachexia.If there is no clear correlation with a clinical benefit (e.g., does reducing CRP lead to increased function), then a biomarker will not fulfil FDA requirements to become a surrogate endpoint.
Overall, the reason behind the lack of trials showing significant improvements in biomarkers in cancer cachexia trials will be multifactorial.The first is a lack of efficacious treatment for cancer cachexia.The second is that many of the trials featured in this review were carried out across multiple tumour types.The underlying mechanisms in cancer cachexia are likely to be different in each cancer type 85 (and perhaps each genotype), and therefore, applying one biomarker or set of biomarkers to test an intervention across multiple cancers is unlikely to show significant results; rather, patient selection may be based on the biomarker.
It is worth noting that in two recent narrative reviews of biomarkers in cancer cachexia, 5,86 both discussed using objective measures of the skeletal muscle wasting process, such as activin A and myostatin.They also suggested the use of GDF-15 and parathyroid hormone-related peptide (PTHrP).Neither of these biomarkers were featured in any of the trials in this review.The recent results of the first study of ponsegromab in participants with cancer cachexia 8 are noteworthy due to it being the first cancer cachexia trial to admit patients based on the biomarker GDF-15.However, given that this trial contained 10 participants, it did not meet the search criteria for this systematic review.The next phase of the study is eagerly anticipated.

Strengths and limitations
The strengths of this review include its prospective design (PROSPERO) and use of broad search criteria encompassing the last three decades of cancer cachexia trials with data extraction performed by multiple independent reviewers.It is important to note the limitations of this review.In an effort to include clinical trials of a high standard, those with a smaller sample size (n < 40) were excluded; however, additional information could have been drawn from these trials and likewise from studies published before 1990.The use of the modified Downs and Black checklist provided a robust assessment in keeping with the other reviews in this endpoint series.Very few studies presented data on the association between the different biomarkers measured.Unless explicitly stated, it is often difficult to tease out whether authors considered outcomes as primary, secondary or exploratory.Therefore, only biomarkers that were explicitly stated as primary outcomes were highlighted in Table 1, and this may underrepresent the perceived significance of some biomarkers.Again, unless explicitly stated, it is difficult to determine if routine biochemistry and haematology were measured in trials as endpoints or as part of the therapeutic monitoring process, and as such, the frequency of these biomarkers as outcomes is likely overstated.
A further limitation of the present review and studies herein was that the utility of a biomarker to predict those most likely to respond to an intervention and/or the reactiveness of a biomarker to the efficacy of an intervention was not assessed.Primarily, this was not only because the studies were rarely powered based on the biomarker being examined but also because the temporal relationship was not frequently described.This would be important to assess in future work, and if paradigms with oncology treatments were realized (e.g., PSA as both a diagnostic and treatment biomarker), this would be an important step for assessing who is most likely to benefit from a cachexia treatment and also to measure effectiveness.

Conclusions
In conclusion, this systematic review of 52 cancer cachexia trials found 99 different biomarkers.Biomarkers are predominately used as exploratory rather than primary endpoints.It seems reasonable that for a biomarker to be responsive to change in the context of a cachexia clinical trial, it must be related to the mechanism of action of the intervention and/or the underlying cachexia process that is modified by the intervention.Further, to reach regulatory approval, the relationship between the biomarker and clinical benefit must be clear.
Foundation, Rosetrees Trust, UKI NETS and NIHR.MJH has consulted for and is a member of the Achilles Therapeutics Scientific Advisory Board and Steering Committee.MJH has received speaker honoraria from Pfizer, Astex Pharmaceuticals, Oslo Cancer Cluster and Bristol Myers Squibb and is a co-inventor on a European patent application relating to methods to detect lung cancer (PCT/US2017/028013).BJAL has received personal fees for consulting from Artelo, Actimed, Faraday, Kyona Kirin and Toray.RJES has received personal fees for consulting from Artelo, Actimed, Faraday and Helsinn.EJR has a consulting/advisory role for Napo, AIM Specialty Health, Oragenics, BASF, Immuneering, Vector Oncology, Asahi Kasei, Heron, Pfizer/EMD Serono and Mitobridge.

Figure 3
Figure 3 Balloon plot demonstrating the relationship between biomarkers and intervention in cancer cachexia trials.Trials yielding statistically significant changes between intervention arms are filled in.CRP, C-reactive protein; GPS, Glasgow prognostic score; Hb, haemoglobin; IGF-1, insulin-like growth factor 1; IL-6, interleukin-6; TNF, tumour necrosis factor.