Sarcopenia as a risk factor of progression-free survival in patients with metastases: a systematic review and meta-analysis

This systematic review was conducted based on a predefined protocol and in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [30] and Meta-analysis of Observational Studies in Epidemiology (MOOSE) recommendations [31]. The review protocol was registered in PROSPERO with the registration unique identifying number (UIN) of CRD42022325910.

Databases of Medline, Embase and Cochrane Central Register of Controlled Trials were systematically searched from inception to October 14, 2022 by two independent investigators (MZ and XS) without language or time restrictions. We used MeSH (for Medline and Cochrane)/Emtree (for Embase) terms combined with free-text words (including synonyms and closely related words) that were associated with metastases and sarcopenia.The detailed search strategy and specific terms were used, which were searched as free text words and as MeSH/Entrée terms without language restrictions. In addition, we also performed manual reference check of relevant articles, meta-analyses, reviews, and meeting abstracts. When two or more articles used the same cohort data, we preferred the most up-to-date ones with full-text information available. We perform study selection by a series of consecutive stages including duplicate checking using Endnote software, titles and abstracts screening, full-text article selection according to the eligibility criteria. These processes were conducted independently by two investigators (XS and TW). Conflicts were handled by consensus, and an adjudicator (WG) was consulted when necessary. If different opinions were encountered, senior experts would be consulted (SF or YF).

Studies were considered appropriate and were included in the analysis if they satisfied the following established inclusion criteria. [1] prospective or retrospective population-based cohort study design; [2] participants: Patients diagnosed with metastatic cancer, with or without sarcopenia. Sarcopenia was diagnosed according to the criteria given by EWGSOP in 2019 [2, 32]; [3] outcome: progression-free survival; [4] the measure of association: hazard ratio (HR) and corresponding 95% CIs provided from the original studies or having related data that could be used to calculate the risk ratios. We excluded hospital-based on community-based observational studies and those providing inadequate data to generate risk ratio for the association between sarcopenia and metastases.

Two investigators (XS and TW) independently read through, screened, extracted data from the included studies, and filled in the pre-designed data extraction excel forms. If there were any discrepancies, we would consult a third senior investigator (WG) until a consensus was reached. The following study characteristics were abstracted including study author, publication year, study design, study period, geographical region, observation period, population characteristics and age at cancer diagnosis, main treatment, measurements and definitions of sarcopenia, original cancer type, metastatic site of cancer and outcome.

The methodological quality for the included studies was evaluated using the Newcastle–Ottawa scale (NOS) [33] tool. It evaluates the cohort study through three modules and eight items in total, specifically, it includes the selection, comparability, and exposure/outcome evaluation of the study.. NOS uses the semi-quantitative principle of the star system to evaluate the quality of literature. Except for the maximum of 2 stars for comparability, the other items can be evaluated up to 1 star, with a full score of 9 stars. The higher the score, the higher the research quality.

All statistical analyses were performed using Stata statistical software (version 15.1). The primary outcome was the PFS, defined as the length of time during and after the treatment of cancer, that a patient lives with the disease but the disease does not get worse. We applied the DerSimonian and Laird random effects meta-analysis to pool HRs along with the corresponding 95% CIs due to the anticipated substantial heterogeneity in terms of the enrolled populations [34]. To meta-analyze the HRs of PFS, we converted reported HRs to log HRs and used a generalized inverse variance method with a random effects model combining data. Results are reported with both effect estimates and 95% CIs. We used the I² statistic to assess heterogeneity between studies, with I² values > 50% indicating significant heterogeneity [35]. To explore the sources of heterogeneity, we carried out a series of subgroup analyses based on geographical regions (Europe and Asia), gender (male or female), original cancer type, variable analysis type of HR value (univariate vs multivariate), and methodological quality (low or high). Sensitivity analysis was performed by applying the leave-one-out method. Trim-and-fill technique is a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. funnel plot was used to detect publication bias in studies reporting overall survival, with a P-value < 0.1 indicating a significant difference [36].

The initial literature search identified a total of 888 citations. After duplication removal, 756 studies remained for title and abstract review. During this process, we excluded 688 irrelevant citations and 68 potentially relevant studies were selected for full-text review. Due to non-population-based cohorts, reviews, meta-analyses or no outcome data reported, 16 studies [27‐29, 37‐49] involving 1,675 participants satisfied the inclusion criteria and were eligible to be included in the final meta-analysis (Fig. 1).

Table 1 presents the baseline characteristics of the included studies. Among the studies published between 20 16 and 2022, 8 studies [27‐29, 37, 40, 42, 45, 46] were from Asia, 7 studies [39, 41, 43, 44, 47‐49] were from Europe and a separate study [38] from Brazil. All of the studies were retrospective cohort, and 68.75% of the included studies (11/16) were of high quality with an NOS score of ≥ 7). The sample size of the included studies ranged from 29 to 353 participants. Original cancer type included hormone-sensitive prostate cancer, renal cell carcinoma, head cancer, lung cancer, renal cell cancer, bladder cancer, colorectal cancer, gastrointestinal stromal tumor, esophago–gastric junction cancer, proximal gastric cancer, distal gastric cancer, thyroid cancer, upper tract urothelial carcinoma, bile duct cancer, ampullary cancer, breast cancer and pancreatic cancer. 14 studies used the L3 skeletal muscle index (L3-SMI) to measure sarcopenia, 1 study used L3 psoas muscle index (L3-PMI) and 1 study used the total psoas area index (TPI). L3-SMI meant measuring the cross-section area of skeletal muscles (cm²) at L3 disc space divided by the square of the height of the patient (m²), the muscles are mainly the psoas major, and also include erector spinae, quadratus lumborum, transversus abdominis, external and internal oblique, and rectus abdominis muscles, for L3-PMI, the muscle is only the psoas; TPI meant measuring the total area of psoas area (cm²) divided by the square of the height of the patient (m²). For groups used L3-SMI, sarcopenia was defined as L3-SMI < 41 cm²/m² in women, < 43 cm²/m² in men with BMI < 25 kg/m², and < 53 cm²/m² in men with BMI > 25 kg/ m²; For group used L3-PMI, sarcopenia was defined as L3-PMI ≤ 6.36 cm²/m² for men and ≤ 3.92 cm²/m² for women; For group used TPI, sarcopenia was defined as TPI < 5.73 cm²/m² for men and < 4.37 cm²/m² for women.

Using the NOS tool for cohort studies, a total of 5 studies [37, 40, 42, 44, 48] had a high risk of bias, with each study having 2 to 3 possible sources of bias, bias was most common in adequacy of follow-up. A total of 9 scores were assigned to 9 item questions, and a score of less than 7 was defined as high risk of bias (Table 2).

When we meta-analyzed the 16 studies, the results showed that the pooled HR of progression-free survival reached 1.56 (95% CI = 1.19–2.03) in all site cancer survivors compared with noncancer controls. Heterogeneity among studies was high (I2 = 76.3%; P < 0.001) (Fig. 2).

This systematic review and meta-analysis of 16 population-based cohort studies demonstrated a statistically significant increased risk of reduced PFS in metastatic cancer patients with concomitant sarcopenia compared with non-sarcopenic patients, with results remaining consistent after adjusting for potential publication bias. Furthermore, our results suggest that sarcopenia has a stronger contribution to worsening PFS in Asian populations, a finding that is stable in high-quality studies. Our findings regarding the association between sarcopenia and the risk of worsening PFS in patients with metastatic cancer are consistent with a systematic review and multiple clinical studies, all of which suggest that patients with sarcopenia have a risk of worsening PFS [48, 51, 52]. However, the results of most of these studies were from cohorts with high limitations. These studies were limited to specific cancer types and cancer patients treated with chemotherapy or radiation (e.g., patients with metastatic renal cell carcinoma treated with cabozantinib). To our knowledge, this study is one of the few studies involving a representative population of multiple cancer types, with meta-analysis and systematic review from high-quality population-based cohort studies, rather than previous individual or narrative studies.

Originally used to describe the loss of muscle mass with age, the European Working Group on Sarcopenia in Older People (EWGSOP) has recently defined sarcopenia to include impaired muscle strength and poor physical performance [2]. Previously, most people considered sarcopenia as an inevitable part of aging. However, the degree of sarcopenia is highly variable and depends on the presence of certain risk factors, such as lack of exercise, age-related decreases in hormone concentrations and cytokine imbalances, decreased ability to synthesize proteins, failure of satellite cell activation, potential effects of microRNA, oxidative stress due to chronic diseases, etc. [53‐57]. There are two broad types of sarcopenias, primary sarcopenia caused by aging and secondary sarcopenia mainly caused by malignancy [58]. In cancer patients, the adverse effects of sarcopenia include increased susceptibility to adverse events, increased complications from cancer surgery and chemotherapy toxicity, and difficulty in pursuing further cancer-directed therapy [59]. Sarcopenia is characterized by a decrease in both muscle strength and mass, and therefore may increase the risk of falls and fractures in older adults [60, 61]. In addition, decreased muscle function can affect swallowing and breathing, which can aggravate or cause respiratory disease [62]. At the same time, we note that sarcopenia may interact with certain underlying diseases in patients, leading to a poor prognosis. In one study, it was confirmed that patients with COPD are often associated with sarcopenia and negatively affect important clinical outcomes [63]. For the heart, sarcopenia is thought to be closely associated with heart failure (HF) [64]。Sarcopenia may promote HF development through different mechanisms, including pathological ergoreflex [64, 65]. At the same time, HF may induce sarcopenia through multiple pathways, such as hormonal changes, poor nutrition, lack of exercise and etc. Therefore, early identification and intervention for the basic diseases in patients with metastatic cancer may be necessary in the future.

It is now generally accepted that chemotherapy is also a cause of sarcopenia in cancer patients [66‐68]. However, the relationship between the two is equally complex. One study found that patients with metastatic colorectal cancer (mCRC) had a significant reduction in muscle area during chemotherapy, and the rate of muscle loss observed in the study was 24 times faster than the normal rate of muscle loss (1% per year) [69]. Another study of adjuvant chemotherapy in patients with colon cancer found that baseline sarcopenia was associated with an increased incidence of all grade 3–4 chemotherapy-induced toxicities [70]. In addition, in patients with colon, lung, esophageal, gastric, and other types of cancer, mortality increases and progression-free survival decreases after chemotherapy, even if patients with sarcopenia do not exhibit lower overall survival [70‐73]. The current explanation for this phenomenon is that patients with sarcopenia are forced to reduce doses or delay dosing cycles due to excessive toxicity to oncology treatment [74]. It is common practice to base chemotherapy on the body surface area of each patient, without taking into account the large and unpredictable fraction of body weight accounted for by adipose tissue [75]. A large amount of evidence shows that this method fails especially in patients with sarcopenia, and the related toxicity risks cannot be solved [76‐78]. Obviously, although some physiological constants may be related to body surface area, they are not related to other anthropometric parameters, such as body mass index. Besides, sarcopenia also reduces the effective effect of chemotherapy. Studies have shown that among breast cancer patients with sarcopenia, many chemotherapeutic drugs such as capecitabine, paclitaxel, docetaxel, and nab-paclitaxel have poor effects [74, 79]. The reasonable explanation may be that the adverse outcome may be related to the high toxicity rate, which in turn may lead to the necessary dose reduction and the provision of effective tumor treatment at a lower dose, thus reducing the therapeutic benefit [80]. To sum up, cancer patients are prone to sarcopenia before and after chemotherapy, and sarcopenia occurring or aggravated during chemotherapy will worsen the prognosis and aggravate the toxicity caused by chemotherapy. Therefore, chemotherapy cycles and doses need to be carefully set based on drug toxicity and therapeutic effects.

In our study, we found that the risk of sarcopenia and reduced PFS was not significant in the European population. It was found that, with the exception of the study of Haik, the remaining studies that showed unstable results were low risk of bias studies, and sensitivity analysis also yielded relatively stable results after excluding this study [44]. All patients in this study were from a single hospital cohort and did not target a specific cancer, but rather included multiple cancers. In addition, this study was the only cohort to use immune checkpoint therapy on patients, which has relatively few side effects. Other studies have commonly used chemotherapy, and patients may have been treated with the previously described, mutually reinforcing effect of sarcopenia and chemotherapy toxicity, but this was not a problem in immunotherapy. This could be a potential reason why this research concluded that sarcopenia was not significantly associated with PFS. Several studies have now demonstrated that cancer patients treated with immune checkpoint inhibitors alone or in combination with chemotherapy have significantly improved survival compared to chemotherapy alone [81‐83]. Whether immunotherapy exerts a better improvement in PFS in patients with sarcopenia needs to be further explored in more clinical studies.

This meta-analysis provided important future clinical implications for the risk of worsening PFS in metastatic cancer patients with concomitant sarcopenia, yielding an approximate risk estimate with an HR of 1.56 (95% CI = 1.19–2.03) for PFS. Early screening and effective interventions are clinically important in the prevention and treatment of sarcopenia. Second, optimal strategies for prevention and management of sarcopenia have not been established due to the widespread neglect of sarcopenia in cancer. Due to the lack of reliable clinical data to guide clinicians, physicians may consider the use of specific treatment options based on the history of adjuvant therapy (e.g., chemotherapy or radiation therapy), baseline information, etc., in cancer patients. In addition, screening and treatment of individuals at high risk for stroke has implications for the prevention of sarcopenia and for reducing the burden of sarcopenia in the general elderly population.

The current study has several advantages in the following aspects. First, the current systematic evaluation and meta-analysis includes a more representative population in the relevant field, providing up-to-date evidence on the association between the risk of worsening PFS in metastatic cancer patients with concomitant sarcopenia. Second, we developed a systematic and comprehensive database search strategy based on the major online databases (Medline, Embase, and Cochrane Library) with no search date restrictions so that we could retrieve as many relevant articles from around the world as possible, avoiding publication bias on pooled results and improving the reproducibility of results. Third, almost all included studies were from national cohorts or population-based cohorts, thus minimizing potential selection bias stemming from study design. In addition, a transparent methodological quality assessment of the included studies was performed using the NOS list for cohort studies. Fourth, several methods, including subgroup analysis and sensitivity analysis, have been applied to thoroughly identify sources of heterogeneity based on abstract study-level baseline characteristics. In these sensitivity analyses, after excluding one low-quality study, our results remained stable and Egger's test combined with the cut-and-patch method found no evidence of publication bias.

This study also has some limitations. First, significant heterogeneity was found in the included studies, which was predictable and may be partly due to differences in baseline characteristics of the population (sex, race, tumor primary site and metastatic site, etc.), exposure to treatment (adjuvant chemotherapy, radiation therapy, or hormonal therapy), and statistical methods (adjustment for confounders). Although several methods were applied to adjust for outcomes, considerable moderate to high heterogeneity remained. After careful study, we concluded that irregularities in PFS can introduce bias in the results, mainly due to differences in the interval between patient follow-up after treatment. For PFS, PFS is susceptible to the influence of the follow-up interval because the exact time point at which a patient progresses is uncertain. Different follow-up intervals may lead to highly variable trial results. For example, if patients are asked to follow up every 3 months, assuming a high degree of patient compliance, and if patients in the trial group have a prolonged survival of less than 3 months compared to the control group, it cannot be concluded that the trial group is better than the control group because tumor progression in both groups will be detected at the same review every 3 months, then PFS is recorded as No difference. In a more extreme case, if the tumor is reviewed every 3 months after treatment, and the tumor is determined to be progression-free at the first post-treatment review (3 months) in the trial group, but progression appears soon after the review (much less than 3 months, or even days) but is not detected, while tumor progression is detected at the first post-treatment review (nearly 3 months) in the control group, the difference in progression-free survival between the two groups is considered to be 3 months, while the true difference between the two groups is close to 0. The true difference is close to 0, which would result in a 3-month bias. Such differences can occur not only within studies, but also between studies due to inconsistent PFS intervals, and the shorter the follow-up interval for patients, the more reliable their trial results. This inference is supported by the fact that most of our studies did not mention the follow-up interval and that those that reported this element were at low risk of bias and had a higher degree of confidence, as other investigators have also mentioned this directly or indirectly [84]. However, given that the results of most subgroup analyses and sensitivity analyses were highly consistent with the primary outcome, we believe that the impact of these heterogeneities on the primary outcome of the study is limited. Second, the results of this meta-analysis are based on observational cohort studies, which may be limited by confounding factors such as patient gender, adjuvant treatment modality, and follow-up interval. However, all included studies provided data comparing sarcopenic and non-sarcopenic populations and matched with a number of important covariates suggesting a consistently increased risk of worsening PFS in sarcopenic patients. Third, because our study was a single-study-level meta-analysis rather than an individual patient-level meta-analysis, we were unable to perform more detailed subgroup analyses (e.g., risk analyses based on events during cancer survival and follow-up) and were unable to explore progressive sarcopenia that occurred during follow-up. Fourth, some subgroup analyses found nonsignificant results, which we believe may be due to the relatively small sample size and low statistical efficacy. More evidence from high-quality prospective cohort studies on the impact of sarcopenia on the risk of death or progression in patients with metastatic cancer is needed. Finally, the meta-analysis was limited to studies published in peer-reviewed journals in English. We may have missed articles published in other languages or in journals outside of the three databases we searched. In addition, unpublished gray literature was not included. However, three major databases, Medline, Embase, and the Cochrane Library, published the vast majority of available reports. Despite these limitations, the current study includes the vast majority of cancer types, which provides a largely adequate sample size for meaningful and robust statistical analyses.