1. Introduction
Cancer remains one of the leading causes of mortality worldwide and is responsible for 8.8 million deaths each year. In westernised countries, it has been estimated that one in three people will develop cancer in their lifetime and one in four will die from it [
1,
2].
The importance of cachexia syndrome, with escalating nutritional and functional decline leading to poor clinical outcomes, is well recognised [
3]. However, how this complex syndrome is best defined is the subject of continuing debate. Clearly, defining any syndrome is difficult due to its multifaceted nature. However, in such circumstances one may resort to the duck test approach: “If it looks like a duck, swims like a duck, quacks like a duck, then it probably is a duck”. Such abductive reasoning has been commonly used to settle legal cases and more recently has gained popularity in artificial intelligence.
In the context of cancer cachexia, a number of factors have been shown to impact independently on quality of life (including functional and symptom scores) and survival. These include Eastern Cooperative Oncology Group-Performance status (ECOG-PS) and the systemic inflammatory response (modified Glasgow Prognostic Score, mGPS), both of which have been extensively validated [
4,
5,
6,
7]. More recently, based on an international consensus, body mass index/weight loss (BMI/WL) grades have been shown to impact independently on quality of life and survival [
8,
9,
10].
Recently, these three criteria for the definition of cancer cachexia were directly compared and all three independently predicted survival in patients with advanced cancer [
11]. However, BMI/WL grade was low risk in approximately 50% of patients and ECOG-PS and mGPS were independently associated with survival in this group. Therefore, to further investigate the clinical utility of these three criteria to define cachexia, the aim of the present study was to examine the relationship between ECOG-PS, mGPS, BMI/WL grade, and Computerised Tomography (CT)-derived body composition and physical function tests in patients with advanced cancer.
3. Results
A total of 523 patients (266 males, 257 females) satisfied the inclusion criteria. The relationship between clinicopathological characteristics, body composition, and physical function is shown in
Table 2. The majority of patients were over 65 (56.8%), had a BMI >25 kg/m2 (50.1%), and had metastasis (83.2%). Gastrointestinal (34.4%) and lung (31.7%) cancers were the most common tumours. The median overall survival was 5.6 months (95% CI: 5.1–6.0 months). At the date of censoring, 318 patients (61%) were dead. Median follow-up time for patients that had died was 10.5 months (95% CI: 9.0–12.1 months).
The relationship between ECOG-PS and measures of body composition and physical function are shown in
Table 3. ECOG-PS was significantly associated with skeletal muscle index (
p < 0.05), skeletal muscle radiodensity (
p < 0.001) and timed up and go (
p < 0.001).
The relationship between mGPS and measures of body composition and physical function are shown in
Table 4. mGPS was significantly associated with skeletal muscle radiodensity (
p < 0.01), timed up and go test failure (
p ≤ 0.001), and hand grip strength test failure (
p < 0.01).
The relationship between BMI/WL grade and measures of body composition and physical function are shown in
Table 5. BMI/WL grade was significantly associated with visceral obesity (
p < 0.05) and skeletal muscle radiodensity (
p < 0.01).
Low skeletal muscle radiodensity was significantly associated with timed up and go test failure (n = 192, p = 0.015) and hand grip strength failure (n = 100, p = 0.042).
The relationship between ECOG-PS, mGPS, BMI/WL grade, and subcutaneous fat index in patients with advanced cancer is shown in
Table 6. On multivariate binary logistic regression analysis, BMI/WL grade (OR 0.62, 95%CI 0.40–0.97,
p < 0.05) remained independently associated with a high subcutaneous fat index.
The relationship between ECOG-PS, mGPS, BMI/WL grade and high visceral obesity is shown in
Table 6. On multivariate binary logistic regression analysis, BMI/WL grade (OR 0.57, 95%CI 0.38–0.87,
p < 0.01) remained independently associated with a high visceral obesity.
The relationship between ECOG-PS, mGPS, BMI/WL grade and low skeletal muscle index is shown in
Table 6. On multivariate binary logistic regression analysis, ECOG-PS (OR 1.90, 95%CI 1.51–2.39,
p < 0.001) remained independently associated with a low skeletal muscle index.
The relationship between ECOG-PS, mGPS, BMI/WL grade and low skeletal muscle radiodensity is shown in
Table 6. On multivariate binary logistic regression analysis, ECOG-PS (OR 1.68, 95%CI 1.11–2.55,
p < 0.05), mGPS (OR 1.32, 95%CI 1.01–1.73,
p < 0.05) and BMI/WL grade (OR 1.50, 95%CI 1.02–2.19,
p < 0.05) remained independently associated with a low skeletal muscle radiodensity.
The relationship between ECOG-PS, mGPS, BMI/WL grade and timed up and go test is shown in
Table 6. On multivariate binary logistic regression analysis, ECOG-PS (OR 5.84, 95%CI 3.79–9.00,
p < 0.001) remained independently associated with timed up and go failure.
The relationship between ECOG-PS, mGPS, BMI/WL grade and hand grip strength is shown in
Table 6. On multivariate binary logistic regression analysis, mGPS (OR 1.95, 95%CI 1.29–2.97,
p < 0.01) remained independently associated with hand grip strength failure.
4. Discussion
Over the last decade or so there has been increasing interest in identifying objective criteria to define cancer cachexia. This has proven problematic since cancer cachexia is a syndrome impacting on quality of life, body composition, physical function, and survival. In the present study, candidate criteria were directly compared in terms of their relationship with measures of body composition and physical function. ECOG-PS and mGPS were consistently associated with low skeletal muscle mass and function and therefore, together with our previous study [
11], both ECOG-PS and mGPS would appear to pass the duck test as criteria to define cancer cachexia.
In the present study, poor performance status was significantly associated with low skeletal muscle index, low skeletal muscle radiodensity, and timed up and go test failure but not hand grip strength test failure. Furthermore, high mGPS was significantly associated with low skeletal muscle radiodensity, timed up and go test failure, and hand grip strength test fail. In contrast, high BMI/WL grade was significantly associated with high subcutaneous fat index, high visceral obesity, and low skeletal muscle radiodensity. Therefore, BMI/WL grade appears to capture elements of the decline in fat mass. The present and previous [
11] results clearly need to be repeated to prove the clinical utility of the ECOG-PS/mGPS framework. However, if this proves to be the case (and these observations are readily repeated) there are a number of important implications for the future diagnosis and treatment of cancer cachexia. The present results would suggest that in addition to ECOG-PS, mGPS is useful in defining the syndrome of cancer cachexia. Therefore, the ECOG/mGPS framework should be considered as part of routine assessment prior to treatment in patients with advanced cancer.
In the present study it was of interest that low skeletal muscle radiodensity was significantly associated with timed up and go test failure (
p < 0.05) and hand grip strength test failure (
p < 0.05). These results would be consistent with the results of a recent study by Williams and co-workers who reported that skeletal muscle radiodensity was related to physical function impairments including activities of daily living (ADL), climbing stairs, walking, and timed up and go [
29]. Furthermore, the presence of systemic inflammatory response degrades the quality of the skeletal muscle [
30]. If this were to be the case then it might be anticipated that downregulation of the systemic inflammatory response, compared with placebo, would result in better preservation of muscle density, muscle strength, and performance status. This hypothesis is the subject of a number of ongoing randomised clinical trials. For example, there is a randomised placebo controlled phase III trial underway of a multimodal intervention (exercise, nutrition, anti-inflammatory medication) in patients with advanced lung or pancreatic cancer undergoing anti-cancer therapy with palliative intent (NCT02330926) [
31]. The aim of this trial is to prevent or attenuate loss of weight, muscle, and physical function using a multimodal intervention which is anti-inflammatory. The findings from the associated phase II trial provide grounds for optimism for the ongoing phase III trial [
32].
In the present study, three criteria were considered to define cachexia. With reference to ECOG-PS this has long been considered a cornerstone of assessment by oncologists and palliative physicians. With the increasing integration of oncology and palliative care this is likely to remain an important part of the assessment of the patient with advanced cancer. It may be that other objective measurements of “real life” performance status will more consistently reflect ECOG-PS, such as activity trackers (e.g., Fitbit) [
33,
34,
35]. With reference to the mGPS there has been in recent years extensive validation of its use in patients with advanced care, and routine assessment is now advocated [
36,
37]. Of the present criteria considered, it is the only one that is completely objective as it relies on two routine, laboratory-derived values. Indeed, it has been termed “laboratory cachexia” as its values become increasingly abnormal towards death [
38] and the mGPS above has been used to define cancer cachexia [
39,
40]. There are other measures of the systemic inflammatory response that have been shown to have prognostic value, such as the neutrophil lymphocyte ratio which can be collected as part of the routine differential white cell count. However, such ratios have not been well defined and their relationship with the syndrome of cachexia has not been shown [
7,
41,
42]. With reference to BMI/WL grade it is not clear whether this has additional value to other nutritional risk screening tools such as the Malnutrition Universal Screening Tool (MUST) and the Patient Generated Subjective Global Assessment (PG-SGA) that are in routine clinical use [
36,
40]. Therefore, further comparative studies are required to establish the value of BMI/WL grade as a measure of cachexia in patients with advanced cancer.
The findings of the present study may also help inform regulatory endpoints in the arena of trials treating cancer cachexia. To date there has been a lack of concordance in regulatory guidance between the EMA and FDA regarding endpoints [
43] whilst previously agreed endpoints of skeletal muscle mass and function have not been realized in multiple clinical trials of varying agents [
44,
45,
46,
47]. It may be that moderating the systemic inflammatory response in patients with advanced cancer will produce more reproducible gains.
Limitations of the present study include that body composition measures and physical function test data were not available in all patients. In the present study the data were analysed according to clinically relevant criteria (previously reported to be associated with clinical outcomes such as survival) rather than statistical criteria. Specifically, categorical rather than continuous analysis was used and since only 10 out of 403 subjects passed the two-minute walk test this was excluded from further analysis. Furthermore, the cohort was relatively heterogenous with different cancer types and specific stages of disease. However, when further stratification of the results was carried out for both lung cancers and gastrointestinal cancers in particular (n ~180 each,
Supplementary Tables S1 and S2), similar results were obtained on univariate and multivariate analysis to that of the combined cohort, suggesting that the relationships between ECOG-PS, mGPS, BMI/WL grade and skeletal muscle index, skeletal muscle density and physical function were not specific to cancer type. With reference to stage of disease, more than 80% of patients had metastatic disease on study entry and therefore the heterogeneity of this cohort would have been unlikely to confound the present results. Importantly, the present results are likely to represent the type of patient cohort being treated by both oncologist and palliative care physicians. Further work is required to define these relationships in specific tumour types and at specific stages of disease. Furthermore, objective ongoing measurements of physical function such as the use of Fitbit monitors would be of considerable interest.