The measurement of skeletal muscle (SM) and adipose tissue (AT) in oncology research has become increasingly relevant in an era of precision medicine. These measures have potential to inform personalized cancer care and treatment planning—increasing safety and identifying those who require additional multidisciplinary care. Computed tomography (CT) scans that are routinely performed in the oncology setting for tumour evaluation have secondary value as an accurate means of body composition measurement. SM and AT cross-sectional areas measured from single magnetic resonance imaging (MRI)/CT images in the lumbar region are linearly related to whole-body SM and AT, both in healthy adults and in cancer patients.1,2 The specifics of this method have been thoroughly described elsewhere.2,3 In short, trained observers select a single axial image slice at the middle of the third lumbar vertebra (L3), which is imported into image analysis software. SM and AT are semi-automatically delineated based on established Hounsfield unit ranges and manually corrected as required.
Several metrics are used to describe SM and AT based on CT analysis at L3; SM and total AT cross-sectional areas at L3 (in cm2) are the raw output of single slice analysis and correlate with whole-body SM mass and whole-body fat mass, respectively, according to published regression equations.1,2 SM area at L3 can be normalized for height as SM index (SMI; cm2/m2) for comparison between people of different heights.4 With respect to body composition change, cross-sectional area (cm2) and index units (cm2/m2) are valid metrics of absolute change, while relative change (%) from baseline is relevant for assessing severity of loss.
Using CT-based analysis of SM and AT, researchers have defined various thresholds of SMI at cancer diagnosis associated with poor prognosis.5–10 The subsequent measurement of direction and intensity of change in SM and AT during cancer treatment may enhance the prediction of treatment outcomes. For most patients with solid tumours, more than one routine CT scan is taken over the cancer trajectory, enabling opportunistic measures of SM and AT over time. Investigators are increasingly measuring SM and AT change during chemotherapy treatment to explore associations with patient outcomes. These analyses have resulted in an abundance of reported data, yet heterogeneity precludes valid meta-analysis, which requires that studies be sufficiently congruent with respect to population, exposure, outcome measure and time interval of measurement.11
Three recent meta-analyses of body composition change during cancer treatment have been published; all report significant heterogeneity among included studies, suggesting uncertainty of the resulting estimates. Jang et al. endeavoured to determine the mean change in SM observed during any chemotherapy regimen (neoadjuvant, adjuvant, curative or palliative) for any cancer type.12 Of the potential 92 studies, the authors excluded 77 because of insufficient reporting (i.e., change was not reported in cm2/m2, or medians and ranges were reported instead of mean and standard deviations). Significant heterogeneity (I2 = 86.83%) was found among included studies, related to variations in cancer type, stage, treatment regimen and treatment duration. The results of this meta-analysis are difficult to apply to any specific patient group, given the inclusion of all disease sites, stages and chemotherapy regimens.
Another meta-analysis was performed by Xu et al. with the aim of describing the prognostic impact of SM change during neoadjuvant treatment for gastrointestinal cancer.13 These authors clearly defined the disease site (gastrointestinal) and setting (neoadjuvant) but found heterogeneity related to population studied, treatment protocol, outcome measurement and reporting. The meta-analysis included studies measuring SM change with bioelectric impedance analysis, psoas muscle on CT and cross-sectional muscle area on CT, and SM change was reported variously in total lean body mass change (kg), L3 psoas or SMI change (cm2/m2), L3 area change (cm2), and using variable classifications of ‘muscle loss’. Finally, the time between measurements in each study was not addressed as a source of heterogeneity.
Finally, Wang et al. conducted a meta-analysis of studies describing SM change and its prognostic value during neoadjuvant therapy for oesophageal and oesophagogastric junction cancers.14 While this protocol had the most specific treatment and disease-site criteria of the three reviews mentioned, the authors chose to process the reported results from each included study to estimate the observed change in SMI units (cm2/m2), as few studies reported these units. The overall estimate of change during neoadjuvant treatment was found to have high heterogeneity, with an I2 value of 88.3%, in part due to treatment regimen. The authors did not describe or discuss the impact of time between measurements on the meta-analysis results.
In summary, despite an abundance of publications describing SM change during cancer treatment, meta-analysis estimates of this change and its impact are uncertain due to heterogeneity in population, disease site, treatment regimen, measurement method and metrics reported. When primary data are still emerging, scoping reviews are a rigorous systematic form of evidence synthesis that can be used to explore the size and extent of existing literature, summarize what is presently known in a general sense, identify under-represented populations and suggest elements of study design and reporting that will enable future meta-analysis.11,15,16 No systematic attempt has been made to synthesize current knowledge on body composition change in the setting of palliative-intent chemotherapy. A preliminary search confirmed that this literature is heterogeneous with respect to population, time interval of measurement and reporting of metrics, like those cited above. We applied scoping review methodology using the palliative-intent setting, to illustrate the heterogeneity impeding reliable meta-analysis of SM and AT change and associated outcomes and to suggest standards to facilitate evidence synthesis. The specific objectives were to (1) demonstrate the methodological variability in measurement and reporting of SM or AT change and their associated outcomes during cancer-directed treatment and (2) propose a strategy with respect to design, reporting and publication standards for studies measuring body composition change, which will enable evidence synthesis.
A preliminary search of PROSPERO, MEDLINE, the Cochrane Database of Systematic Reviews and JBI Evidence Synthesis was conducted and no current or in-progress scoping reviews or systematic reviews on the topic were identified.
MethodsThis scoping review followed a pre-defined protocol, registered on 5 April 2022 on Open Science Framework (registration DOI:
Studies from any geographical setting were included if they (1) evaluated adult patients (≥18 years old) during receipt of palliative-intent chemotherapy for a solid tumour and (2) measured change in SM and/or AT during palliative-intent chemotherapy, without intervention intended to attenuate SM and AT loss, by analysis of axial CT images at the L3 vertebra.
Exclusions were intended to narrow the scope of the review to demonstrate that even within a defined setting, CT-based analysis of SM and AT change is not standardized. Studies were excluded if ≥25% of patients in the sample were receiving neoadjuvant therapy, radiotherapy, surgical resection, or exclusive targeted therapy or immunotherapy. Retrospective and prospective observational designs were considered, along with studies describing a standard care control group of a clinical trial.
Search strategyThe search aimed to locate both published and unpublished studies (such as theses) and is reported according to the PRISMA-ScR extension of the PRISMA (2020) guidelines.17 An initial limited search of MEDLINE (1946 to present via Ovid) was undertaken to identify index terms used to describe relevant articles, which were used to develop a full search strategy for MEDLINE (Ovid) (see supporting information); it employed both controlled vocabularies, such as MeSH and EMTREE, and keywords representing key concepts. The search strategy, including all identified keywords and index terms, was adapted for each database, and reference lists of articles selected for full-text review were used to screen for additional papers. MEDLINE (1946 to present via Ovid), CINAHL Plus with Full Text (EBSCO), Embase (1974 to present via Ovid), Web of Science-All Databases (Clarivate Analytics), which in itself includes Web of Science Core Collection, BIOSIS Citation Index, BIOSIS Previews, CABI: CAB Abstracts, Derwent Innovations Index, KCI-Korean Journal Database, Russian Science Citation Index, SciELO Citation Index and Zoological Record, and Cochrane Library (Wiley Online Library) were searched. Cochrane Library (Wiley Version) was also searched independently. Sources of unpublished studies and grey literature included Dissertations and Theses Global (ProQuest) and websites such as beta.asco.org,
Following the search, all identified records were uploaded into Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia) and duplicates removed. Titles and abstracts were screened by two independent reviewers (S. S. and P. K.) for assessment against the inclusion criteria. Abstracts that did not reflect the inclusion criteria were not reviewed in full text. Potentially relevant papers were retrieved in full and were assessed against the inclusion/exclusion criteria by two independent reviewers (S. S. and P. K.). Reasons for exclusion of full-text papers are reported in Figure 1. Data were extracted by two independent reviewers (S. S. and P. K.) using a data extraction tool developed by the reviewers using Covidence software. The data extraction template was registered with the protocol and revised once during the extraction process to create additional space for reported body composition metrics. Any disagreements that arose between the reviewers during the process were resolved through discussion or by a third reviewer (V. M.).
Enlarge this image.Figure 1. PRISMA flow diagram detailing the selection of sources of evidence. CT, computed tomography.
The systematic search identified 2869 publications, containing 372 duplicates. A total of 2496 publications were screened by abstract and title, with 2413 deemed irrelevant based on inclusion and exclusion criteria. Full text was reviewed for 83 publications, with 47 excluded for reasons related to (1) setting (neoadjuvant, adjuvant, curative-intent or surgical treatment) or (2) outcome measurement (tissue change was not measured or described; single time point only; not during chemotherapy; or not measured using CT). A total of 38 publications were included for data extraction (Figure 1). Of these, five were re-analyses of the Dutch Colorectal Cancer Group CAIRO3 cohort, for which the design and description of longitudinal body composition change was primarily reported by Kurk et al.18 The remaining four CAIRO3 analyses19–22 were considered secondary analyses of associations with body composition change as previously reported. These were considered collectively as ‘CAIRO3’ to describe cohort characteristics, study design and reporting of body composition change, while each analysis was considered individually for the purpose of describing associations between body composition change and clinical outcomes.
Characteristics of included cohortsCharacteristics of the 34 unique cohorts representing a total of 3933 patients are summarized in Table 1. The majority of patients included were from Asia (43%), followed by Europe (41%) and North America (16%). Colorectal cancer was the most represented disease site, followed by pancreas cancer and cholangiocarcinoma. Advanced genitourinary and breast cancers represented 2.9% and 3.6% of the total sample, respectively (Figure 2). By design, disease sites for which monitoring does not include CT scans and/or the L3 level, such as head and neck cancers and haematological malignancies, were not represented.
Table 1 Characteristics of 34 cohorts in which body composition change was measured during palliative-intent chemotherapy
| Author, year | Disease site | Region | Male (%) | Majority chemotherapy regimen | Multiple regimens (Y/N) | |
| Rier, 201823 | Breast | Netherlands | 98 | 0 | Paclitaxel or 5-FU + DOX + cyclophosphamide | No |
| Solomayer, 201924 | Breast | Germany | 29 | 0 | NR | — |
| Antoun, 201925 | Colorectal | France | 76 | 50 | XELIRI/FOLFIRI ± bevacizumab | No |
| Best, 202126 | Colorectal | USA | 226 | 53 | FOLFOX ± bevacizumab | Yes |
| Blauwhoff-Buskermolen, 201627 | Colorectal | Netherlands | 63 | 63 | CAPOX ± bevacizumab | Yes |
| CAIRO318–22 | Colorectal | Netherlands | 450 | 63 | CAP(OX) ± bevacizumab | No |
| Dolly, 202028 | Colorectal | France | 72 | 63 | FOLFIRI | Yes |
| Gallois, 202129 | Colorectal | France | 137 | 55 | FOLFOX or FOLFIRI ± bevacizumab | Yes |
| Maddalena, 202130 | Colorectal | Italy | 56 | 59 | NR | — |
| Malik, 202131 | Colorectal | Poland | 78 | 55 | Trifluridine + tipiracil hydrochloride | No |
| Sasaki, 201932 | Colorectal | Japan | 219 | 65 | FOLFOX ± bevacizumab | Yes |
| Van der Werf, 202033 | Colorectal | Netherlands | 54 | 55 | CAPOX ± bevacizumab | Yes |
| De Jong, 201934 | Lung | Netherlands | 116 | 55 | Platinum + paclitaxel + bevacizumab | No |
| Kakinuma, 201835 | Lung | Japan | 44 | 71 | Multiple | Yes |
| Lee, 202136 | Lung | Korea | 70 | 89 | Platinum + gemcitabine | Yes |
| Murphy, 201037 | Lung | Canada | 41 | 46 | Platinum + vinorelbine | Yes |
| Murphy, 201138 | Lung | Canada | 24 | 50 | Platinum-based doublet | Yes |
| Naito, 201739 | Lung | Japan | 30 | 63 | Platinum-based doublet | Yes |
| Stene, 201540 | Lung | Netherlands | 35 | 51 | Platinum + gemcitabine | Yes |
| Birgitte Stene, 201941 | Lung/pancreas | UK/Norway | 46 | 57 | NR | — |
| Babic, 201942 | Pancreas | USA | 164 | 58 | Gemcitabine-based or FOLFIRINOX | Yes |
| Basile, 201943 | Pancreas | Italy | 94 | 55 | Gemcitabine-based or FOLFIRINOX | Yes |
| Choi, 201544 | Pancreas | Korea | 484 | 61 | Gemcitabine | Yes |
| Salinas-Miranda, 202145 | Pancreas | Canada | 105 | 61 | FOLFIRINOX | Yes |
| Kays, 201846 | Pancreas | USA | 53 | 62 | FOLFIRINOX | No |
| Uemura, 202147 | Pancreas | Japan | 69 | 55 | FOLFIRINOX | No |
| Rollins, 201648 | Pancreas/cholangio | UK | 98 | 56 | Gemcitabine-based | Yes |
| Cho, 201749 | Cholangio | Korea | 524 | 66 | Gemcitabine + platinum | Yes |
| Dijksterhuis, 201950 | Gastro-oesophageal | Netherlands | 65 | 75 | CAPOX | Yes |
| Feng, 202051 | Gastric | China | 46 | 63 | Epirubicin + oxaliplatin + fluorouracil | No |
| Park, 202052 | Gastric | Korea | 111 | 72 | FOLFOX or CAPOX | Yes |
| Rimini, 202153 | Gastric | Italy | 40 | 60 | FOLFOX | Yes |
| Fukushima, 201854 | Genitourinary | Japan | 72 | 74 | Gemcitabine + platinum | Yes |
| Nagai, 201955 | Genitourinary | Korea | 44 | NR | Gemcitabine + docetaxel | No |
Abbreviations: 5-FU, 5-fluorouracil; CAPOX, capecitabine, oxaliplatin; DOX, doxorubicin; FOLFIRI, leucovorin, 5-fluorouracil, irinotecan; FOLFIRINOX, leucovorin, 5-fluorouracil, irinotecan, oxaliplatin; FOLFOX, leucovorin, 5-fluorouracil, oxaliplatin; NR, not reported.
Enlarge this image.Figure 2. (A) Disease site and (B) geographical distribution of 3933 patients from 34 cohorts in which body composition change was measured during palliative-intent chemotherapy. GE, gastroesophageal; GU, genitourinary.
The chemotherapy regimen of majority for each study is shown in Table 1. Majority regimens were multi-agent in 30/34 studies, single agent in one study and not reported in three instances. Eight studies (25.8%) were limited to patients on a single regimen, or presented results disaggregated by regimen, while the remainder (74.2%) included patients on more than one regimen and presented aggregate results.
ChallengesPresentation of the aggregate data of patients on multiple therapy regimens limits exploration of regimen-specific changes in SM and AT during treatment. Palliative chemotherapy regimens vary within disease sites, and thus, even meta-analysis by disease site would result in heterogeneity due to regimens used, as previously identified.12–14 Recent narrative reviews suggest regimen-specific changes in SM56–58; this can only be confirmed with systematic review and meta-analysis of studies that report SM and AT change by regimen. If reports indicate that tissue change does not differ by regimen within a disease site, similar regimens could be legitimately aggregated in subsequent work.
Proposed standard
- Report regimen-specific changes in SM and AT, concurrent with presentation of overall changes.
Study designs with respect to the timing and reporting of body composition measurement are graphically summarized in Figure 3, ordered from most specific to least specific timing of baseline scan. Baseline scan was reported as occurring prior to palliative chemotherapy in most designs (25/34); of these, 12 clearly specified a time frame. The baseline scan was described as ‘at diagnosis’ or ‘first CT’ or ‘staging’ in six designs.24,26,36,46,48,53 In three instances, the baseline scan occurred within a range of days before or after treatment initiation.38,43,52
Enlarge this image.Figure 3. Variation in study design and reporting of skeletal muscle (SM) and adipose tissue (AT) change during palliative-intent chemotherapy. Scan interval: reported median or amean days between scans. Checkmark: metric was reported. Dash: metric was not reported. %: relative SM change. cm2: change in cross-sectional area. cm2/m2: change in skeletal muscle index. Cut-off: the proportion of patients who reached an SM change cut-off of interest. By sex: sex-specific reporting. CT, computed tomography scan; d, days; Dx, diagnosis; IMAT, intramuscular AT; Measure. error, measurement or precision error; mo, months; NA, not applicable; SAT, subcutaneous AT; TAT, total AT; VAT, visceral AT.
Endpoint scan timing was similarly variable, defined in seven instances by a single time point (e.g., 6 weeks) after treatment start.25,26,29,33,45,47,59 Five studies defined endpoint as a window of time (e.g., 60–120 days) after treatment start.28,32,37,39,42 Seven designs defined endpoint CT timing using a set number of chemotherapy cycles, ranging from one to six cycles; however, the length of one cycle was not often reported.23,34,40,50,51,54,55 Ten designs specified a treatment or disease milestone as the endpoint, including first reassessment,30,31,43,53 progression18,44,49,52 and last CT.36,46 Non-specific descriptors for the endpoint, such as ‘during treatment’ or ‘after treatment’, were applied in five designs.24,27,35,38,48 The actual mean or median time between scans was reported for 15/34 cohorts, ranging from 64 to 362 days.
ChallengesThe timing of baseline and endpoint measurement provides context for the interpretation of change in SM and AT. Lack of a clearly defined time frame for the pre-chemotherapy baseline scan means that scans taken long prior to palliative-intent chemotherapy may be included, and not account for change that may have primarily occurred prior to the current treatment. Similarly, endpoint delineation based on disease or treatment milestones represents the greatest challenge to interpretation and reproducibility of results; this is particularly true when last CT or CT indicating progression is the endpoint, given inter-individual variation in time to these milestones. Finally, infrequent reporting of actual scan interval impedes interpretation of the presented data, particularly when baseline and endpoint timing is not clearly defined.
Recognizing that any prospective or retrospective review of imaging acquired during standard of care will result in variable CT scan timing, clear definition of the time over which change is measured must be a priority. For example, ‘the CT scan within 45 days prior to palliative chemotherapy initiation was selected as the baseline; if multiple scans were available in this window, the closest to treatment initiation was used,’ and ‘the CT scan within 90–120 days after chemotherapy initiation was selected as the endpoint scan; if multiple scans were available in this window, the closest to 120 days was used.’
Proposed standard
- Study protocols must clearly describe inclusion criteria for baseline and endpoint CT scans in units of time from a common reference point.
In all included studies, measurements of SM and AT were based on the total cross-sectional areas at L3, rather than specific muscles such as psoas only. The measurement error of CT analysis for SM was reported in eight instances. Five reports referenced measurement error from prior literature, either 2% error18,28,37,40 or 1.3% error34 (Figure 3). Four reports provided a calculated inter-observer coefficient of variation for SM measurements, ranging from 0.6% to 2.4%.21,27,33,42 Of these, one also differentiated between intra-observer variability (difference in repeated measures by the same observer) and inter-observer variability (difference between observers).21
ChallengesThe concept of measurement error of CT analysis has not been acknowledged by most investigators. Among included studies, the precision error of observers on repeated measures was only described in one report. Measurement of precision error enables classification of patients with true tissue loss, stable tissue or true tissue gain according to a least significant change value. Further, considering the least significant change when defining scan interval ensures that the interval is long enough to allow for detection of changes beyond measurement error.
Proposed standard
- The least significant change must be reported, as determined by precision error testing following a published method for repeated measures such as that described by Arribas et al.60
At baseline, SM status was described using at least one metric for 33/34 cohorts, using SMI (cm2/m2) in 22/33 instances and SM area (cm2) in 12/33. Total body muscle mass (kg) was estimated using regression equations in three instances. Thirteen publications included sex-specific reporting of baseline SM. Two cohorts were entirely female; thus, reporting by sex was not applicable. Baseline AT and weight metrics were reported less often than SM. Eleven publications contained at least one baseline AT metric, reported by sex in three instances. Visceral, subcutaneous and total AT areas (cm2) or indices (cm2/m2) were used variably. Intermuscular AT and estimated fat mass (kg) were each reported in one publication. Baseline body mass index (kg/m2) was presented for 22 cohorts, and in seven instances, this was reported categorically. Baseline mean weight (kg) was reported in four instances.
The metrics used to report SM change in each study are visualized in Figure 3. Six publications did not quantify the change observed, but rather reported the proportion of patients who reached a particular cut-off of interest for SM loss. Metrics used to describe mean/median SM change included cm2/m2 (16/34), % (13/34) and cm2 (11/34). SM change was comprehensively described with three metrics in two publications (2/34), two metrics in eight publications (8/34) and one metric in the remainder (18/34). Eight publications normalized SM change to a specific time period, ranging from 30 days to 1 year. Eight publications disaggregated SM change by sex, while the remainder pooled males and females.
AT changes were described for 10 cohorts, with two disaggregated by sex. In one instance, total AT change was described singularly, while the remaining nine specified subcutaneous, visceral and/or intermuscular AT change. Weight change was described for 16/34 cohorts using variable metrics including kg, % or kg/m2.
ChallengesReporting of SM and AT at baseline is necessary to contextualize the sample population and the changes observed. Sex-specific reporting of baseline SM and AT is uncommon, despite differing central tendencies and distributions of SM area and SMI between males and females.61
Reporting of SM and AT change has been limited to one or two metrics, and in some cases, not even described, particularly if the primary outcome is related to a pre-defined cut-off of SM or AT loss. AT changes are less commonly reported than SM, representing a gap in the literature. Normalization of observed change (i.e., to a standard number of days) has been applied in data sets with widely variable scan intervals, which assumes that the rate of body composition change is constant, even in the last days of life, or that it can be extrapolated from short scan intervals. This method introduces uncertainty and estimation to the reported data.
Finally, reporting of sex-specific changes in SM and AT is rare. Muscle mass and biological characteristics of muscle are known to be different between males and females.61 The distribution, function and behaviour of AT are also divergent between males and females.62–64 Whether change in each of these tissues over time is uniform or whether it differs between males and females remains to be characterized and will only be determined if outcomes are reported by sex, which is rarely done. At present, this remains an unrecognized potential source of variation in studies evaluating longitudinal changes in SM and AT.
Proposed standards
- Report baseline SM and AT in cm2 and cm2/m2, by sex.
- Clearly report actual time between scans (mean/median and range of days) for each analysis to allow the reader to interpret the observed changes in the context of time. Normalization of change over a standard time period is not an alternative to clearly defining baseline and endpoint scan inclusion criteria.
- Report changes in SM and AT by sex using multiple metrics, including both absolute and relative changes, using supplementary materials if required. Consider including a waterfall plot to visualize the distribution and central tendency of change for each tissue.40,65
Survival was the most frequent outcome investigated in relation to SM change (n = 24), with 22 publications including clearly presented Cox proportional hazards models using SM and/or AT change as an independent variable. A summary of the models used for survival analyses is presented in Table 2. In most instances, change was measured within the first 100 days of palliative treatment.
Table 2 Summary of Cox proportional hazards models evaluating the independent association between skeletal muscle change during palliative chemotherapy and overall survival
| Reference | Scan interval | Classification of SM change for model | Rationale for classification | Model covariates | |
| 23 | 6 cycles | SM loss > 0 cm2 | Unspecified | 98 | Age, ER/PR positive; number of metastases; stage at diagnosis, regimen |
| 25 | 64 days | ‘SM score’, not clearly described | Unspecified | 57 | Age, sex, metastases, regimen, BMI |
| 26 | 89 days | SM loss > 5% | Miyamoto et al. (2015) | 193 | Age, sex, tumour mutational status, weight loss, total AT loss |
| 27 | 78 days | SM loss > 9% | Tertiles | 67 | Sex, age, LDH, comorbidity, metastases, treatment line, response |
| 20 | To progression | Continuous SMI change (per standard deviation) | NA | 450 | Age, sex, PS, stage, primary site, resection, initial disease response, LDH, metastases, dose reduction, scan interval |
| 29 | 60 days | SM loss > 14% | Log-rank maximization | 149 | Hypoalbuminemia, nutrition risk score, response |
| 30 | To first reassessment | SM loss > 5% | Miyamoto et al. (2015) | 56 | Model not presented |
| 31 | 105 days | SM loss ≥ 5% | Unspecified | 78 | Histological differentiation, CEA |
| 32 | 60–120 days | SM loss > 9% | Blauwhoff-Buskermolen et al. (2016) | 142 | Age, sex, BMI, PS, prior resection |
| 36 | First to last CT | Top tertile rate of SM loss, cm2/30 days | Tertiles | 70 | Age, stage, PS, disease response |
| 40 | 88 days | SM loss > 2% | Mourtzakis et al. (2008) | 35 | Sex, PS, stage, response, quality of life and appetite loss at baseline, BMI, regimen |
| 42 | 80 days |
(a) Top quartile SM loss (b) Top quartile AT loss |
Sex-specific quartiles | 164 | Age, study site, race, baseline SM/AT, sex, year, stage, BMI, diabetes, smoking, regimen |
| 43 | To first reassessment | SM loss ≥ 10% | Sugiyama et al. (2017) | 94 | Tumour stage, visceral AT, PS change |
| 44 | To progression | SM loss > 2 cm2/m2 | Unspecified | 484 | Age, sex, PS, disease extent, BMI, sarcopenia, BMI change, best response |
| 45 | 77 days | Continuous SMI change (per cm2/m2/30 days) | Log-rank maximization | 105 | Disease response |
| 46 | First to last CT | SM loss > 5% plus AT loss > 5% | Unspecified | 53 | Age, sex, disease extent, response, sarcopenia, obesity, sarcopenic obesity, myosteatosis, tumour location |
| 47 | 71 days |
(a) SM loss ≥ 7.9% (b) AT loss ≥ 5.4% |
Median | 69 | Age, sex, metastases, jaundice, obstruction, diabetes, tumour size/location, CA19-9, CEA, BMI, albumin, UGT1A1 heterozygous, response, AT index, AT change, SMI, SMI change, sarcopenia |
| 49 | To progression | SM loss > 7% | Unspecified | 524 | Age, sex, primary site, PS, regimen, stage, SMI, BMI, change in BMI, response |
| 50 | 79 days | Continuous SMI change (per cm2/m2) | NA | 65 | Age, sex, PS, metastatic sites |
| 51 | 2 cycles | SM loss > 8% and/or VAT loss > 20% | Quartiles | 46 | Model not presented |
| 52 | To progression | SM loss > 6.5 cm2/m2 | Tertiles | 111 | Sarcopenia at baseline (Korean specific), PS, overall response rate |
| 53 | To first reassessment | SM loss > 10% | Unspecified | 40 | PS |
| 54 | 2 cycles | Continuous SMI change (per cm2/m2) | NA | 72 | Disease sub-type, C-reactive protein |
| 55 | First to last CT | Rate of SM loss > 1%/30 days | Unspecified | 44 | Age, sex, PS change, comorbidities, response, HGB, weight change, metastatic site, time |
Abbreviations: AT, adipose tissue; BMI, body mass index; CA19-9, cancer antigen 19-9; CEA, carcinoembryonic antigen; CT, computed tomography; ER/PR, oestrogen receptor/progesterone receptor; HGB, haemoglobin; LDH, lactate dehydrogenase; NA, not applicable; PS, performance status; SM, skeletal muscle; SMI, SM index; VAT, visceral AT.
Continuous SMI change was used as a predictor of survival in three studies, while most specified a cut-off for SM loss to create a dichotomous variable using % (n = 14), cm2 (n = 3), cm2/m2 (n = 2) or unspecified (n = 1). Commonly applied cut-offs were 5% SM loss (n = 4), 9% SM loss (n = 2) and 10% SM loss (n = 2). Cut-offs were applied with no stated rationale in eight instances or selected based on the sample data in six instances (i.e., median, tertiles or quartiles). Five authors referenced prior literature when applying cut-offs, and two defined cut-offs based on the log-rank maximization method.
Among 22 presented Cox proportional hazards models, covariates applied included age, sex, disease response, disease stage/status (e.g., locally advanced and metastatic), metastatic spread (e.g., location and number), baseline body mass index or weight, biological values (e.g., albumin), treatment type, tumour-specific factors (e.g., mutational status, size and primary location), comorbidities, baseline SM, baseline AT and time. The prevalence of use for each of these covariates is visualized in Figure 4. Three models included co-occurring AT change as a covariate.26,46,47 No sample size assumptions or calculations were presented for multivariable survival analyses; however, several authors noted inadequate sample size as a limitation.
Enlarge this image.Figure 4. Prevalence of covariate inclusion among 22 Cox proportional hazards survival models evaluating the association between skeletal muscle or adipose change and survival in patients with advanced cancer. BMI, body mass index.
The relationship between SM change and chemotherapy toxicity was explored in eight studies. Of these, five publications compared the incidence of treatment toxicities between ‘SM losers’ and ‘SM non-losers’ (i.e., univariable),25,27,31,32,40 and three included SM loss in multivariable logistic regression to predict odds of toxicity.22,29,50 The relationship between SM loss and changes in health-related quality of life or physical function was explored in four studies.19,39,40,59
ChallengesMultiple definitions of SM loss have been evaluated for association with survival, ranging from any loss to 14% loss, over variable periods of time. Rationale for the selection of these cut-offs is rarely provided. Given the wide range of cut-offs used to categorize SM loss published in the literature for survival prognostication, studies defining a new cut-off for categorization of the independent variable in survival analysis should be well powered and clearly indicate where the selected cut-off falls within the distribution of observed change. The use of a continuous independent variable (SM change and/or AT change) in a survival model should be done with consideration of precision error. For example, if measured least significant change is 2.0 cm2, a model using continuous SM loss per cm2 may be unreliable. The ideal interval for evaluation of SM or AT change to inform prognostication will vary based on the expected survival of the cohort; however, prognostication based on early change is the most feasible due to attrition.
Survival analyses are often underpowered and presented as exploratory, thus are at high risk of overfitting.66 The prognostic impact of CT-defined SM loss may be related to concurrent disease progression,40,67 yet several survival models did not account for disease response. Co-occurring total adipose loss is rarely considered as a covariate, even though AT loss and SM loss together represent the main components of total tissue loss, or a comprehensive view of habitus change over time.
In studies evaluating the relationship between SM loss and toxicity, inconsistent definitions of treatment toxicity have been used. Further, the time of SM change measurement often occurred during the period in which toxicity was evaluated, making it impossible to ascertain exposure versus outcome. Finally, the relationship between SM loss and patient-reported outcomes is a clear gap in the literature that will require prospective studies. As data availability in small disease sites may preclude adequate power, standardized design and reporting, data repository deposit, and/or multi-centre collaboration will strengthen meta-analysis of prognostic models.
Proposed standards
- Measure change over a consistent time period for all patients in the sample.
- Provide clear rationale if using a single tissue change cut-off for prognostication, and ensure the cut-off is greater than the least significant change.
- Include known covariates (age and sex) and account for concurrent changes (disease response and total adipose change) in survival models.
- Clearly differentiate between the period in which tissue loss occurs and the period in which the risk of toxicity is evaluated when attempting to define a causal relationship between SM change and treatment toxicity. For example, ‘this analysis evaluated the association between SM loss > 5% in the first 3 months of chemotherapy and the risk of toxicity in the subsequent 2 months of chemotherapy.’
Since Mourtzakis et al. presented CT image analysis as a ‘practical and precise approach to quantification of body composition measurement in cancer’,2 this method has been applied worldwide to measure body composition status and change during cancer treatment. CT-defined SM and AT change have been measured across multiple cancer sites and treatment plans, limited only when abdominal CT scans are not routinely used for monitoring. A variety of landmarks have emerged and been validated for measurement in regions other than L3, and the same challenges apply.60 Despite an abundance of data, reliable meta-analysis describing change in body composition during palliative chemotherapy, and its related outcomes, remains challenging. Using systematic scoping review methodology, literature from a single cancer treatment setting was used to demonstrate barriers to meta-analysis and to propose a minimum standard for future reports. Our results demonstrate wide variability within and between studies related to treatment protocols, scan intervals (time over which change is measured), change metrics reported and the treatment of SM and AT change in prognostic models.
The setting of palliative-intent chemotherapy was used as a sample to narrow the scope of literature, which may reduce the generalizability of our findings to other settings such as curative-intent cancer treatment or non-cancer settings. Measurement of SM and AT from CT scans taken as part of standard oncological care naturally leads to inconsistent measurement timing, as clinicians rather than researchers select the timing of CT evaluations. This issue is particularly accentuated in palliative settings, where patients who decline rapidly may have early re-evaluation of CT scans due to worsening status, while others are re-evaluated according to a standard schedule. Defining and reporting the boundaries of inclusion for baseline and endpoint scans is increasingly important in groups such as this, where scan intervals are variable and influenced by the clinical imperatives for repeat scanning.
Other challenges identified by this review are not specific to the use of routinely acquired CT scans for SM and AT measurement. While prospectively planned studies or those using dual-energy X-ray absorptiometry (DEXA), bioelectrical impedance analysis (BIA) or another technique with the sole purpose of longitudinal body composition measurement will have greater control over measurement timing, it remains imperative for authors to report measurement error, baseline/endpoint timing and the actual interval between measurements.
Regardless of the setting, key principles for study design, measurement, reporting and statistical analysis should be applied to future reports (Figure 5). Consensus discussions would further enhance these recommendations to produce publication standards, enabling accurate meta-analysis of body composition change and associated clinical outcomes. Meta-analysis of carefully designed and clearly reported measurements of SM and AT change will move this body of research towards meaningful application of findings to clinical care.
Enlarge this image.Figure 5. Principles for measurement and reporting of computed tomography (CT)-defined skeletal muscle (SM) and adipose tissue (AT) change and associated outcomes in patients receiving cancer-directed treatment.
The authors wish to thank Dr. Vickie Baracos and Dr. Lisa Martin for their helpful review and comments. PK is supported by the Canadian Institutes of Health Research in the form of a Canada Graduate Scholarship—Doctoral. The authors certify that they comply with the ethical guidelines for authorship and publishing in the Journal of Cachexia, Sarcopenia and Muscle.68
Conflict of interest statementThe authors declare no conflicts of interest.
© 2023. This work is published under http://creativecommons.org/licenses/by-nc-nd/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Thorlakson, Jessica 2 ; Servais, Stephane 3 1 Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, AB, Canada
2 University of Alberta Libraries, University of Alberta, Edmonton, AB, Canada
3 Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, AB, Canada; Faculté de Médecine, Inserm UMR1069 Nutrition Croissance et Cancer, Université de Tours, Tours Cedex, France