Pre-Therapeutic Sarcopenia among Cancer Patients: An Up-to-Date Meta-Analysis of Prevalence and Predictive Value during Cancer Treatment

This study will address the prevalence of pre-therapeutic sarcopenia (PS) and its clinical impact during cancer treatment among adult cancer patients ≥ 18 years of age. A meta-analysis (MA) with random-effect models was performed via a MEDLINE systematic review, according to the PRISMA statement, focusing on articles published before February 2022 that reported observational studies and clinical trials on the prevalence of PS and the following outcomes: overall survival (OS), progression-free survival (PFS), post-operative complications (POC), toxicities (TOX), and nosocomial infections (NI). A total of 65,936 patients (mean age: 45.7–85 y) with various cancer sites and extensions and various treatment modes were included. Mainly defined by CT scan-based loss of muscle mass only, the pooled prevalence of PS was 38.0%. The pooled relative risks were 1.97, 1.76, 2.70, 1.47, and 1.76 for OS, PFS, POC, TOX, and NI, respectively (moderate-to-high heterogeneity, I2: 58–85%). Consensus-based algorithm definitions of sarcopenia, integrating low muscle mass and low levels of muscular strength and/or physical performance, lowered the prevalence (22%) and heterogeneity (I2 < 50%). They also increased the predictive values with RRs ranging from 2.31 (OS) to 3.52 (POC). PS among cancer patients is prevalent and strongly associated with poor outcomes during cancer treatment, especially when considering a consensus-based algorithm approach.


Introduction
Since its first definition, introduced in 1989 as an age-associated loss of muscle mass, the definition of sarcopenia has been gradually refined [1]. In 2010, 2011, and 2014, the European Working Group On Sarcopenia (namely EWGOS 1), the International Working Group on Sarcopenia (IWGS), and the Asian Working Group on Sarcopenia (AWGS 1) agreed to define sarcopenia as a syndrome characterized by an age-related loss of skeletal muscle mass (quantitatively assessed by the skeletal muscle index (SMI) or appendicular skeletal muscle mass (ASM) using CT scan, and bioelectrical impedance analysis (BIA) or dual-energy X-ray absorptiometry (DXA) methods), and by loss of function (loss of muscle strength and/or physical performance) using a screening-based approach targeting gait speed measurement [2][3][4]. In 2019, the AWGS updated the threshold values of the operational criteria (then termed AWGS 2). The EWGOS definition was also updated (EWGOS 2), now defining sarcopenia as a muscle disease characterized by the association of low levels of muscle strength (handgrip strength) and low muscle mass, with low physical performance (typically slow gait speed) becoming an indicator of severity [5,6]. In addition, the condition does not affect solely older adults ≥65 years and it has now been recognized that sarcopenia can begin earlier in life. In particular, sarcopenia is considered as primary or age-related, and as secondary when a specific cause (mainly driven by inflammatory processes) is evidenced (Table A1).
Cancer is frequently considered to be a major cause of secondary sarcopenia. In our previous systematic review including 35 observational studies or clinical trials and 6894 patients with cancer before 2016, sarcopenia concerned 38.6% of patients before cancer treatment [7]. We found that pre-therapeutic sarcopenia (PS) was associated with poor survival rates, post-operative complications and chemotherapy-related toxicities during cancer treatment, but with a strong between-study heterogeneity with respect to cancer site or extension and the definitions of sarcopenia. Since this review, with the arrival of immune-therapies and an update in the consensuses, there has been a considerable increase in the number of additional studies among cancer patients, but no comprehensive analysis has been conducted to date.
We therefore aimed to update our previous systematic review and to decipher heterogeneity in the prevalence of PS and its predictive values for overall survival, progressionfree survival, post-operative complications, treatment-related toxicities, disability, and nosocomial infections among cancer patients using a meta-analysis including research published before 2022.

Materials and Methods
We followed the recommendations of the preferred reporting items for systematic reviews and meta-analyses (PRISMA) method for reporting this systematic review with meta-analysis [8]. The protocol was registered on 6 February 2023 and is available on the OSF platform: https://doi.org/10.17605/OSF.IO/H7PUZ, accessed on 13 February 2023.

Information Sources
This meta-analysis was based on a systematic, comprehensive search on MEDLINE via PubMed for articles published in English or French from 31 March 2016 to 31 December 2021. Due to a considerable increase in the number of studies addressing sarcopenia among cancer patients, we chose only to consult the PubMed database. The following research algorithm was used: sarcopenia AND (cancer OR tumors OR malignancies) AND (death OR overall survival OR progression-free survival OR relapse OR chemotherapy OR targeted therapy OR radiotherapy OR hormonal therapy OR surgery OR immunotherapy OR toxicity OR disability OR infection) AND human NOT review NOT letter. All articles retrieved from our previous systematic review were also included [7].

Search Strategy
For this meta-analysis, the following issues were addressed: (a) What is the most commonly encountered definition of PS among patients with cancer? (b) What is the pooled prevalence of PS among patients with cancer, and what is the prevalence according to the definition of sarcopenia? (c) What are the mean differences in muscle strength (i.e., grip-strength) and physical performance (i.e., gait speed) between sarcopenic and non-sarcopenic groups of patients with cancer? (d) What is the predictive value of PS for overall survival (OS) and progression-free survival (PFS) among patients with cancer? (e) What is the predictive value of PS for severe post-operative complications (POC) among patients with cancer? (f) What is the predictive value of PS for severe treatment-related toxicities and/or dose-limiting toxicities (TOX) among patients with cancer? (g) What is the predictive value of PS for disability and nosocomial infections (NI) among patients with cancer?
To answer these questions, we pre-defined eligibility criteria for the articles: patients (adults 18 y and over with cancer), intervention (pre-therapeutic sarcopenia assessed using a consensual measurement), comparator (sarcopenia vs. no sarcopenia), outcomes (prevalence of sarcopenia, OS and PFS, severe post-operative complications defined as a Clavien-Dindo scale score ≥ 3a, ≥grade 3 treatment-related toxicities (CTCAE) and/or dose-limiting toxicities, disability defined in terms of an activities of daily living score (ADL) of ≤5/6, and nosocomial infections defined as hospital-acquired infections), and study design (clinical trials, prospective or retrospective studies with consecutive inclusions) (PICOS) criteria.

Selection Process
Articles meeting the eligibility criteria were first selected on the basis of titles and abstracts (FP) then on the basis of perusal of the full text by 5 independent groups (PBR/MF, ALC/CM, SM/RG, FP/JP, EL/ZapT, and EP/AN/MP). The term sarcopenia was to be clearly defined in the articles. If several articles reported similar results, only the article with the most complete information was retained. Duplicates were screened for and removed. Disagreements were resolved by consensus in each reviewing group.

Data Collection
The data recorded included publication date, country, study design, follow-up time, number of patients, number of men and women, cancer site, cancer extension (classified as local, locally advanced, or metastatic), treatment modes, mean or median age at inclusion, the definition of sarcopenia used (low muscle mass quantity only or consensus-based algorithm), cut-off values for quantitative muscle mass indices (arm muscle area (AMA, cm 2 ), ASM (kg/m 2 ), psoas muscle index (PMI, cm 2 /m 2 ), SMI (cm 2 /m 2 ) or total psoas area (TPA, cm 2 )), muscle strength assessed by handgrip-strength (kg), physical performance assessed by gait speed (m/s), number of sarcopenic patients, number of sarcopenic men and women, number of sarcopenic patients with a body mass index ≥30 kg/m 2 (i.e., sarcopenic obesity), and finally the outcomes associated with either the PS values (%) or the hazard ratios or the odds ratios.

Quality Assessment
We used the Newcastle-Ottawa quality assessment scale (NOS) designed for cohort studies which was the case for all patients, even for those recruited from RCTs [9]. Based on a risk-of-bias assessment, this scale rates the quality of studies with scores ranging from 0 to 9. The quality of the studies was classified as good (≥7), fair (4-6), or poor (0-3).

Effect Measures
The prevalence of PS was summarized as a pooled prevalence with 95% confidence interval (95% CI) using logit transformation.
Handgrip strength and gait speed were summarized as a pooled mean difference (MD) with 95% CI with reference to non-sarcopenic patients using the inverse variance method.
OS and PFS were summarized as a pooled risk ratio (RR) with 95% CI with reference to non-sarcopenic patients using the inverse variance method.
The remaining outcomes were summarized as a pooled RR with 95% CI with reference to non-sarcopenic patients using the Mantel-Haenszel method.

Synthesis Method
The data were analyzed using R statistical software (version 4.1.0; R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org, accessed on 1 September 2022). All tests were 2-sided and statistical significance was set at p < 0.05.
Regarding the study characteristics, categorical variables were summarized as the numbers (%), and continuous variables were summarized as the means ± standard deviation (SD) or medians [Q1-Q3] as appropriate. The studies were described in descending order according to their publication date.
To detect a non-linear relationship between sarcopenia prevalence and the muscle mass indices (using reported cut-off values), we used a non-parametric regression via smoothing splines when possible.
On the basis of the selected articles, and given that between-study heterogeneity was expected, we performed a meta-analysis with random-effect models (with the package "meta") to assess the prevalence of PS, the mean difference in muscle strength and physical performance indices (grip strength and gait speed), and the predictive value of PS for OS, PFS, Clavien-Dindo scale ≥grade 3 for POC, ≥grade-3 for TOX, disability (ADL score ≤ 5/6), and NI among cancer patients. As there was a single study addressing disability, we did not conduct a meta-analysis on this outcome. With regard to the prevalence of PS (first endpoint), we first ran a funnel plot to detect graphical asymmetry. Statistically, the funnel plot asymmetry was assessed using the Peters' test, which is appropriate for meta-analyses of single proportion. We addressed the heterogeneity of the study results using the I 2 indicator and the Cochran's Q test. I 2 values of 0%, 25%, 50%, and 75% were considered to indicate none, low, moderate, and high heterogeneity, respectively. A p value ≤ 0.05 in the Q test indicated a significant heterogeneity. Due to the heterogeneous nature of sarcopenia and the variety of the contexts assessed, we anticipated the need for subgroup analyses according to a sensitivity analysis, excluding studies over the 95% confidence interval from the funnel plot, according to the following: study quality (good, fair or poor), the mean or median age at inclusion classified as < or ≥65 y, sex, BMI (< or ≥30 kg/m 2 ), cancer site, cancer extension, treatment mode, definition of sarcopenia (low muscle mass quantity only or consensus-based algorithm), and the cut-off values for muscle mass indices. We also considered post hoc subgroup analyses according to the publication date (2008-2012, 2013-2017, and 2018-2022), the number of patients included (<100, , and ≥400), and world regions (Asian vs. non-Asian).
To decipher the factors that could explain heterogeneity, we then ran a multivariate meta-regression with a mixed-effect model for the first endpoint (prevalence). Factors (study groups) yielding p values under 0.20 in the univariate analysis were considered for inclusion in the multivariate analysis. A backward selection process of the highest p values was performed to retain the final multivariate model.

Definition of Sarcopenia among Cancer Patients
Sarcopenia was mainly defined from muscle mass measurement only (190/226 studies, 84%), from CT scan (n = 178), BIA (n = 11), or DXA (n = 1) ( Table 1). Sixteen studies did not specify the muscle mass index used [43,45,68,85,86,95,96,120,122,123,139,141,147,148,186,207]. Of the 210 remaining studies, regardless of the sarcopenia definition used, the SMI by CT scan at lumbar three was the main muscle mass quantity index used (171/210, 81 From 2015 to the present, of the 36 studies applying a consensus-based algorithm definition of sarcopenia, 17, 9, 6, and 2 studies respectively used the AWGS1, the EWGOS2, the EWGOS1, and the AWGS2 guidelines. The IWGS was not used.

PS Is Prevalent among Cancer Patients
All the studies were used to assess the prevalence. The pooled prevalence of PS among cancer patients was 38.0% (95% CI: 36.0-41.0) with a high between-study heterogeneity (I 2 = 97%) ( Table 2). Figure A2 shows a significant funnel plot asymmetry (p < 0.0001).    Bold = grouping data, and significant p value at the threshold of 5%; BIA = bioelectrical impedance analysis; DXA = dual energy X-ray sbsorptiometry; AWGS = Asian Working Group on Sarcopenia; EWGOS = European Working Group On Sarcopenia; AMA = arm muscle area; ASM = appendicular skeletal muscle mass; SMI = skeletal muscle index; PMI = psoas muscle index; TPA = total psoas area.
Although the sensitivity analysis excluding studies over the 95% confidence interval from the funnel plot (n = 137 studies) led to less heterogeneity (I 2 = 66%), the 40.5% prevalence was not significantly different (p = 0.11).
The prevalence was significantly lower (p < 0.01) for consensus-based algorithm definitions of sarcopenia (22.0%) than for definitions based on muscle mass measures only (42.0%).
Using the muscle mass measurement-based definition only, the prevalence differed significantly (p < 0.01), ranging from 12.0% (AMA) to 40.0% (SMI). Above the SMI medians of 38.5 cm 2 /m 2 and 47.3 cm 2 /m 2 for women and men, respectively, the prevalence was 47.0% and 52.0%. Figure 1 shows the smoothing splines for the relationship between the prevalence of sarcopenia and muscle mass index measures for women and men (SMI, ASM, and PMI). Regarding SMI, up to the third quartiles of 41.0 cm 2 /m 2 and 52.4 cm 2 /m 2 for women and men, respectively, the association was linear with a tight confidence interval. Regarding ASM, the association was paradoxical for both women and men. For PMI, the association was strictly linear for women and men, but it resulted in a large confidence interval.
Nutrients 2023, 15, x FOR PEER REVIEW 1 the pre-therapeutic prevalence of sarcopenia varied significantly from 33.0% (surge 68% (intra-arterial infusion for hepatocellular carcinoma). In the 26 studies that rep prevalence according to BMI, the prevalence of sarcopenic obesity was significantly (p < 0.01) (19.0%) than for non-sarcopenic obesity (39.0%). The prevalence did not significantly with study quality, the year of publication, the world region, the age-th old of 65 years at inclusion, or sex. In multivariate meta-regression, consensus-based algorithm definitions of sar nia (as opposed to loss of muscle mass only), a sample size ≥ 400, and SMI based scan cut-off values for women (not for men) as continuous variables were independ and significantly associated with the prevalence results (Table A2).  Depending on the cancer site, the prevalence varied very significantly from 24.0% (gastric) to 79.0% (small cell lung). In relation to cancer extension, the prevalence varied significantly from 39.0% (local) to 46.0% (metastatic). According to the treatment mode, the pre-therapeutic prevalence of sarcopenia varied significantly from 33.0% (surgery) to 68% (intra-arterial infusion for hepatocellular carcinoma). In the 26 studies that reported prevalence according to BMI, the prevalence of sarcopenic obesity was significantly lower (p < 0.01) (19.0%) than for non-sarcopenic obesity (39.0%). The prevalence did not differ significantly with study quality, the year of publication, the world region, the age-threshold of 65 years at inclusion, or sex.
In multivariate meta-regression, consensus-based algorithm definitions of sarcopenia (as opposed to loss of muscle mass only), a sample size ≥ 400, and SMI based on CT scan cut-off values for women (not for men) as continuous variables were independently and significantly associated with the prevalence results (Table A2).

Muscle Strength and Physical Performance among Cancer Patients with Sarcopenia
Twelve studies including 3466 patients provided data on muscle strength and physical performance. Handgrip strength values among sarcopenic patients ranged from 17.7 to 22.6 kg and gait speed values among sarcopenic patients ranged from 0.72 to 1.00 m/s [26,51,58,111,138,145,169,177,179,187,206,216]. Figure 2 summarizes the mean differences in handgrip strength and gait speed between sarcopenic and non-sarcopenic patients.  [26,51,58,111,138,145,169,177,179,187,206,216] For handgrip strength, the pooled mean difference was −8.62 kg with a high betweenstudy heterogeneity (I 2 = 91%). Regarding gait speed, the pooled mean difference was -0.19 m/s with a moderate between-study heterogeneity (I 2 = 68%). For handgrip strength, the pooled mean difference was −8.62 kg with a high betweenstudy heterogeneity (I 2 = 91%). Regarding gait speed, the pooled mean difference was −0.19 m/s with a moderate between-study heterogeneity (I 2 = 68%).

Pre-Therapeutic Sarcopenia Is Associated with OS and PFS among Cancer Patients
Based on 101 studies including 28,995 patients, we found a strong, significant association between pre-therapeutic sarcopenia and OS with a pooled RR of 1.97 [1.79-2.17] and with a high between-study heterogeneity (I 2 = 85%, p < 0.01). Subgroup analyses are presented in Table 3 and Figure A3, showing a reduction in heterogeneity. The effect measure differed significantly (p < 0.01) according to sample size, world region, cancer site, and muscle mass index, while no significant differences were found for sensitivity analysis, study quality, year of publication, age threshold of 65 years at inclusion, cancer extension, treatment mode, and definition of sarcopenia. When low between-study heterogeneity was envisaged (i.e., I 2 < 50%), the greatest effects were associated with the PMI-based muscle index (   Bold = grouping data, and significant p value at the threshold of 5%; BIA = bioelectrical impedance analysis; DXA = dual energy X-ray absorptiometry; AMA = arm muscle area; ASM = appendicular skeletal muscle mass; SMI = skeletal muscle index; PMI = psoas muscle index; TPA = total psoas area. For 29 studies including 6546 patients, we found a strong and significant association between pre-therapeutic sarcopenia and PFS with a pooled RR of 1.76 [1.44-2.16] and a high between-study heterogeneity (I 2 = 85%, p < 0.01). Subgroup analyses are presented in Table 4 and Figure A4, showing a reduction in heterogeneity. The effect measure differed significantly (p < 0.01) according to the world region, cancer site, sarcopenia definition, and muscle mass index used, while no significant differences were found for sensitivity analysis, study quality, year of publication, sample size, age threshold of 65 years at inclusion, cancer extension, or treatment mode. When low between-study heterogeneity was envisaged (i.e., I 2 < 50%), the most marked effects were associated with the consensus-based algorithm definitions of sarcopenia (RR = 3.59 [2.17-5.92]) and non-small cell lung cancer (RR = 2.43 [1.90-3.12]).   Bold = grouping data, and significant p value at the threshold of 5%; BIA = bioelectrical impedance analysis; DXA = dual energy X-ray absorptiometry; AMA = arm muscle area; ASM = appendicular skeletal muscle mass; SMI = skeletal muscle index; PMI = psoas muscle index; TPA = total psoas area.

Pre-Therapeutic Sarcopenia Is Predictive of Severe Postoperative Complications among Cancer Patients
Based on 56 studies including 17,172 patients, we found a strong and significant association between pre-therapeutic sarcopenia and severe post-operative complications, with a pooled RR of 2.70 [2.33-3.12] involving a moderate heterogeneity (I 2 = 72%). Subgroup analyses are presented in Table 5 and Figure A5, showing a reduction in heterogeneity. The effect measure differed significantly (p < 0.01) according to sensitivity analysis, year of publication, sample size, world region, cancer site, and sarcopenia definition, while no significant differences were found for study quality, the age threshold of 65 years at inclusion, cancer extension, or muscle mass index. When low between-study heterogeneity was envisaged (i.e., I 2 < 50%), the most marked effects were associated with the consensus-based algorithm definitions of sarcopenia (RR = 3.62 [2.79-4.69]) and gastric cancer (RR = 3.09 [2.42-3.93]).   Bold = grouping data, and significant p value at the threshold of 5%; BIA = bioelectrical impedance analysis; DXA = dual energy X-ray absorptiometry; AMA = arm muscle area; ASM = appendicular skeletal muscle mass; SMI = skeletal muscle index; PMI = psoas muscle index; TPA = total psoas area.

Pre-Therapeutic Sarcopenia Is Predictive of Severe Treatment-Related Toxicity and/or Dose-Limiting Toxicity among Cancer Patients
Based on 19 studies including 2980 patients, we found a significant association between pre-therapeutic sarcopenia and severe treatment-related toxicities and/or dose-limiting toxicities, with a pooled RR of 1.47 [1.17-1.84] involving a moderate heterogeneity (I 2 = 71%). Subgroup analyses are presented in Table 6 and Figure A6, showing a reduction in heterogeneity. The effect measure differed significantly (p < 0.01) according to study quality, sample size, cancer site, cancer extension, and definition of sarcopenia, while no significant differences were found for sensitivity analysis, year of publication, world region, age threshold of 65 years at inclusion, or treatment mode. When low between-study heterogeneity was envisaged (i.e., I 2 < 50%), the most marked effects were associated with breast cancer (RR = 2.93 [1.82-4.73     Bold = grouping data, and significant p value at the threshold of 5%; BIA = bioelectrical impedance analysis; DXA = dual energy X-ray absorptiometry; AMA = arm muscle area; ASM = appendicular skeletal muscle mass; SMI = skeletal muscle index; PMI = psoas muscle index; TPA = total psoas area.

Pre-Therapeutic Sarcopenia Is Associated with Disability among Cancer Patients
Only one study including 131 patients was found on the association between pretherapeutic sarcopenia and disability (ADL ≤ 5/6) [160]. In this single-center prospective study including 40.5% of patients aged ≥ 75 years with cancers in various sites and with different extensions, baseline disability was noted for 30.5% of the patients. Compared to normal muscle mass and non-severe sarcopenia, severe sarcopenia is defined according to the EWGOS1 by low muscle mass (CT scan-based SMI), and both low handgrip strength and slow gait speed were significantly (p < 0.001) associated with disability (90% vs. 26% of patients) in univariate analysis.

Pre-Therapeutic Sarcopenia Is Predictive of Nosocomial Infections among Cancer Patients
Based on 22 studies including 6246 patients, we found a strong, significant association between pre-therapeutic sarcopenia and nosocomial infections with a pooled RR of 1.76 [1.41-2.22] and moderate heterogeneity (I 2 = 58%). Subgroup analyses are presented in Table 7 and Figure A7, showing a reduction in heterogeneity. The effect measure differed significantly (p < 0.01) according to sensitivity analysis, sample size, age threshold of 65 years at inclusion, cancer site, and definition of sarcopenia, while no significant differences were found for study quality, year of publication, world region, cancer extension, treatment mode, or muscle mass index. When low between-study heterogeneity was envisaged (i.e., I 2 < 50%), the most marked effects were associated with gastric cancer (RR = 2.55 [1.88-3.46]) and a sample size ≥ 400 (RR = 2.26 [1.66-3.07]).

Discussion
In this meta-analysis including 226 articles and 65,936 patients with various cancers, various extensions, and various treatment modes, PS was mainly defined as a loss of muscle mass using the SMI on CT scan-based assessment. PS was highly prevalent and was strongly associated with OS, PFS, POC, TOX, and NI during cancer treatment, with pooled relative risks ranging from 1.50 (toxicities) to 2.70 (post-operative complications).
To date, and despite successive sarcopenia consensus-based definitions of sarcopenia provided since 2010, the definition of sarcopenia mainly relies only on loss of muscle mass quantity among cancer patients. The standardized use of CT scans in pre-therapeutic oncological settings probably explains this. Unlike the ASM and the PMI indices, the SMI muscle mass index was linearly associated with the prevalence of sarcopenia for both women and men and had the tightest confidence interval, suggesting that it is probably the most suitable index for the quantification of muscle mass. However, homogeneous optimal cut-off thresholds remain to be clarified in the oncological setting.
As expected, we found great heterogeneity for all endpoints addressed here. Consistent with our previous review, the pooled prevalence of pre-therapeutic sarcopenia concerned 38% of cancer patients [7]. Using a multivariate meta-regression, we were able to identify sources of between-study heterogeneity as follows: consensus-based algorithm definitions of sarcopenia (as opposed to loss of muscle mass only), a powerful sample size (≥400), and the cut-off values of CT scan-based SMI for women (not for men) as continuous variables were independently and significantly associated with the prevalence results for pre-therapeutic sarcopenia. With respect to the prevalence results according to cancer localisation, our results require caution given the impact of the definition used. Strikingly, compared with definitions based on loss of muscle mass only (SMI), consensus-based algorithm definitions of sarcopenia reduced the prevalence significantly (42% vs. 22%), decreasing heterogeneity and increasing the predictive value for OS (RR = 1. . This discrepancy could be explained by the additional criteria used for consensus algorithms, which consider both loss of muscle strength and/or physical performance and muscle mass. Indeed, it is known that grip strength and gait speed are independent factors associated with survival among cancer patients [236]. Surprisingly, except for nosocomial infections, we did not identify any significant difference for prevalence of sarcopenia, OS, PFS, post-operative complications, or severe treatment-related toxicities according to the age threshold of 65. This result highlights the leading role played by cancer (mainly due to cancer-related inflammatory processes) rather than age alone in promoting sarcopenia (namely secondary sarcopenia) and its clinical impact on adverse outcomes [6].
To our knowledge, although it was not performed on individual data, this is the largest and most powerful meta-analysis on this topic. It contains a stringent methodology, bringing together oncologists, geriatricians, and methodologists using data from many countries in numerous cancer settings, enabling us to provide a comprehensive up-to-date review of sarcopenia prevalence and its clinical impact in the course of cancer treatment. In particular, in a cancer setting we were able to highlight an association between pretherapeutic sarcopenia and PFS on the one hand, and between pre-therapeutic sarcopenia and nosocomial infections on the other, subjects that have been studied infrequently to date. However, there are still insufficient data to provide a synthesis regarding the association between pre-therapeutic sarcopenia and disability.
On the basis of the findings of our meta-analysis, there clearly is an urgent need to agree on an operational definition of sarcopenia in oncological settings to improve study comparability. Given both the high prevalence and the strong clinical impact of pre-therapeutic sarcopenia during cancer treatment, we suggest that its detection should occur as early as possible. In agreement with the EWGOS 2 consensus, we support the use of the simple SARC-F (strength, assistance with walking, rise from a chair, climb stairs, and falls) screening tool, which has been previously validated in older cancer patients [237]. The early detection of sarcopenia can help to initiate early muscle rehabilitation combining protein supplementation and resistance exercise training in order to improve the healthcare trajectories during cancer treatment [238]. Finally, we encourage the use of sarcopenia as a stratification variable in the development of future clinical trial designs in oncology.

Conclusions
Using the findings of the largest and the most powerful meta-analysis on this topic to date, we conclude that pre-therapeutic sarcopenia among cancer patients, mainly defined as a loss of muscle mass quantity, is prevalent and strongly associated with OS, PFS, severe post-operative complications, severe treatment-related toxicities and/or dose-limiting toxicities, and nosocomial infections. We stress the need to agree on a consensual definition of sarcopenia in oncological settings.

Data Availability Statement:
The data presented in this study are available upon request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest. Table A1. Consensus-based algorithm definition of sarcopenia.