Adult patients diagnosed with advanced NSCLC at the Department of Oncology of Shangjin Nanfu Hospital, Sichuan University from August 2017 to May 2019 were prospectively and consecutively recruited. Adult patients who met the following criteria were included in the study: (1) pathologically confirmed stage IIIB or IV NSCLC based on the Union for International Cancer Control’s tumor–node–metastasis stage classification and (2) administration of first-line chemotherapy for the first time. However, patients (1) who received immune checkpoint inhibitor therapy, molecular targeted therapy, radiotherapy, or single-agent chemotherapy; (2) who had an implanted pacemaker; (3) who had a history of any other cancer type; (4) who had low-quality CT images or had any anatomical distortion (e.g. chest wall edema) or loss of any muscle mass area on CT images; and (5) who had visible edema were excluded from analysis.

The following clinical variables were recorded by trained nurses within 48 hours upon first admission to our department: age, sex, smoking status (never smoker or ever smoker), histologic type (adenocarcinoma, squamous cell carcinoma, or large cell carcinoma), cancer stage (stage IIIB or IV), Eastern Cooperative Oncology Group (ECOG) performance status (PS) score, number of chemotherapy courses, and chemotherapy regimens. Additionally, serum creatinine, serum albumin, and hemoglobin levels were measured within 24 hours after admission. The updated version of the Charlson comorbidity index (CCI) was used to evaluate the number and severity of important comorbidities, including chronic obstructive pulmonary disease, renal disease, any malignancy, and cerebrovascular disease (20). Data on comorbidities were directly collected from original medical records. The total CCI score is 24 points, whereas the score for ‘any malignancy’ is 2 points (20). Hence, a CCI score ≥3 indicated that a patient not only had NSCLC but also at least one other important comorbidity. Body height and weight were measured using standard methods. Body mass index (BMI) was calculated as body weight (kg) divided by body height squared (m²) and was categorized as follows: underweight, <20.0 kg/m²; normal, 20.0-24.9 kg/m²; and obese, ≥25 kg/m² (4).

Chest CT scans were completed within 48 hours after admission for each participant using a 16-slice spiral CT scanner (Brilliance; Philips Healthcare, OH, USA) with a 5-mm slice thickness. Acquisition parameters were as follows: 100-140 kV, variable mA based on the patient’s body size, and detector collimation of 0.75-1.5 mm. Unenhanced cross-sectional CT images at the T12 level were analyzed using a dedicated segmentation software (Mimics version 21.0; Materialise, Leuven, Belgium) to evaluate the T12 SMA.

On a single CT image, all visualized skeletal muscles with a threshold of −29 to +150 HU, including the erector spinae, latissimus dorsi, rectus abdominis, obliquus externus, internus abdominis, and internal and external intercostal muscles, were segmented (12). The T12 SMI (cm²/m²) was calculated as the T12 SMA (cm²) divided by the body height squared (m²).

A trained observer (L.T.) who was blinded to patient outcomes during the analysis period segmented all CT images. To test the reliability of the T12 SMA determined by CT, a total of 30 participants were randomly selected from the cohort. To assess inter-observer reproducibility, another trained observer (S.H.) subsequently segmented the CT images again. Representative images are presented in eFigure 1 .

On the day of the CT scan, trained nurses utilized a segmental multifrequency BIA device (InBody 770; Biospace, Korea) to measure each participant’s body composition, including the total lean body mass (LBM), trunk LBM, and appendicular LBM.

On the day of the CT scan, trained nurses also measured the handgrip strength to the nearest 0.1 kg using a digital grip dynamometer (EH101; Xiangshan Inc., Guangdong, China) in accordance with the recommendation of the Chinese National Physical Fitness Evaluation Standard (21). Three readings were obtained for each hand, and the highest value was recorded for analysis. Handgrip weakness was defined as handgrip strength <28 kg for men and <18 kg for women (11).

Sarcopenia was defined in the present study as CT-defined sarcopenia and Asian Working Group for Sarcopenia (AWGS)-defined sarcopenia.

(1) CT-defined sarcopenia (based on low SMM estimated by CT image analysis): As there currently exists no established cut-off value of the T12 SMI for determining low SMM, we set the optimal cut-off value for low SMM to predict overall survival (OS) as the diagnostic cut-off points for CT-defined sarcopenia using the maximally selected rank statistics method, as described by Lausen and Schumacher (22). This is a validated method for determining cut-off points that could optimally separate participants with respect to time to an event outcome (4, 23).

(2) AWGS-defined sarcopenia: According to the AWGS 2019 recommendation (11), patients with both low SMM and handgrip weakness (or low physical performance) are considered to have AWGS-defined sarcopenia, whereas patients with low SMM, handgrip weakness, and low physical performance are deemed to have AWGS-defined severe sarcopenia. The PS score is commonly used in routine clinical practice, with a PS score of 0-1 indicating good physical performance (24). In this study, low physical performance was defined as an ECOG PS score ≥2 points.

OS was defined as the number of months elapsed from the date of initial recruitment to the date of death or last follow-up for each patient. Patients were followed up until their death or up to the last week of August 2020, at which time they were confirmed to be alive via telephone review and were subsequently censored.

Data analysis was conducted between October 3, 2020, and October 20, 2020. Histograms and the Shapiro–Wilk test were used to assess the distributions of continuous variables. All continuous variables were of normal distribution. Thus, continuous and categorical variables are expressed as mean (standard deviation) and number (percentage), respectively. Inter-observer validation of T12 SMI measurements was performed using interclass correlation coefficient analysis. Pearson’s correlation coefficient (r) and scatterplots with a linear regression model were employed to assess the association of the T12 SMI with total LBM, trunk LBM, appendicular LBM, BMI, handgrip strength, and the T12 SMA. Hazard ratios (HRs) of the T12 SMI and handgrip strength for predicting OS were evaluated using a Cox proportional hazards regression model with a restricted cubic spline function with three knots for men and women. Subsequently, we determined the cut-off values of the T12 SMI using the maximally selected rank statistics method (22). Using these cut-off values and the diagnostic criteria mentioned above, we defined patients with or without CT-defined sarcopenia and patients with or without AWGS-defined sarcopenia. Group differences were analyzed using one-way analysis of variance and chi-square test (or Fisher’s exact test), as appropriate.

OS curves for different groups were constructed using the Kaplan–Meier method and compared using the log-rank test. Additionally, univariate and multivariable analyses of OS were performed using Cox proportional hazards models, with the results presented as HRs with 95% confidence intervals (CI). Multivariable Cox Models 1, 2, and 3 evaluated the predictive value of CT-defined sarcopenia, AWGS-defined sarcopenia (as well as AWGS-defined severe sarcopenia), and physical performance, respectively, for predicting OS. In line with previous scientific literature (23, 25), the models were adjusted for age at baseline, sex, smoking status, histologic type, cancer stage, CCI, BMI groups, chemotherapy regimens, completion of at least four chemotherapy courses, and serum creatinine, serum albumin, and hemoglobin levels. Moreover, C-indexes were used to assess the discrimination performance of these models to predict OS (25), with c statistics of 0.5 indicating chance; 0.5-0.7, poor discrimination; 0.7-0.8, acceptable discrimination; 0.8-0.9, excellent discrimination; 0.9-0.99, outstanding discrimination; and 1.0, perfect prediction (23).

To assess the robustness of the results, a sensitivity analysis was conducted using the lowest quartile of the sex-specified T12 SMI to define low SMM, which was another widely employed method in previous studies (4). Subsequently, we accordingly redetermined CT-defined sarcopenia, AWGS-defined sarcopenia, and AWGS-defined severe sarcopenia, reperformed the univariate and multivariable analyses with the Cox proportional hazards models, and redrew the Kaplan–Meier curves.

Additionally, a priori subgroup analyses of multivariable Cox Models 1 and 2 were conducted according to age groups (<60 or ≥60 years), sex, cancer stage, histologic type, physical performance (0-1 points or 2 points), and number of chemotherapy courses (1-3 courses or ≥4 courses).

Statistical analyses were performed using R software version 3.5.3 (R Foundation for Statistical Computing, Vienna, Austria) and SPSS software version 26.0 (IBM Corp., Armonk, NY, USA). P values <.05 were considered to indicate statistical significance.

A total of 787 consecutive patients with advanced NSCLC were admitted to our department from August 2017 to May 2019. Of these patients, 82 refused to participate in this study. Furthermore, 15 patients with chest wall edema, 5 patients with low-quality CT images, 16 patients with missing data on handgrip strength, and 30 patients who received molecular targeted therapy were excluded. Accordingly, we included a total of 639 patients (410 men and 229 women) in the study. The median follow-up was 25 months (range: 15-36 months).

The mean age of the participants was 58.6 ± 8.9 years, with men being significantly older than women (mean age: 59.1 vs. 57.7 years, P=.049). The baseline characteristics of participants according to sex are summarized in Table 1 . Overall, 294 (46%) and 345 (54.0%) patients had stage IIIB NSCLC and stage IV NSCLC, respectively. Because the histologic type, chemotherapy regimens, handgrip strength, and body composition variables were significantly different between men and women, we either adjusted for sex or split the data based on sex in the following analyses.

T12 SMI measurements were highly reproducible between observers ( eFigure 2 in the Supplement ). Considering that the T12 SMI is not a classical surrogate of muscle mass, we analyzed the association of the T12 SMI with BIA-derived LBM, handgrip strength, and physical performance. As shown in eFigure 3 in the Supplement , the T12 SMI was highly correlated with trunk LBM and handgrip strength (r=0.78, P<.001 and r=0.70, P<.001, respectively) and was moderately correlated with total LBM (r=0.57, P<.001). Scatterplots with linear regression for these variables are presented in eFigures 4A–E in the Supplement . Moreover, both male and female patients with low physical performance (PS score ≥2) had a significantly lower T12 SMI ( eFigure 4F in the Supplement ).

The T12 SMI was a significant factor for OS in men (P<.001, Figure 1A ) and women (P=.005, Figure 1B ). Handgrip strength was also a significant factor for OS in men (P<.001, Figure 1C ). Furthermore, lower handgrip strength exhibited a tendency toward poor prognosis in women (P=.068, Figure 1D ).

Based on maximally selected rank statistics calculation, we set the cut-off values of the T12 SMI as 32.48 cm²/m² for men and 27.82 cm²/m² for women ( Figure 2 ). At baseline, 160 (25.0%), 141 (22.1%), and 42 (6.6%) patients had CT-defined sarcopenia, AWGS-defined sarcopenia, and AWGS-defined severe sarcopenia, respectively. The prevalence of CT-defined sarcopenia, AWGS-defined sarcopenia, and AWGS-defined severe sarcopenia were not significantly different between men and women ( Table 1 ). The clinicopathological characteristics of participants according to CT-defined sarcopenia or AWGS-defined sarcopenia are summarized in eTable 1 in the Supplement .

A total of 488 (76.4%) patients died during the study period. Patients with CT-defined sarcopenia had a shorter OS than patients without CT-defined sarcopenia ( Figure 3A , log-rank P<.001). Similarly, patients exhibiting handgrip weakness had a shorter OS than those with normal handgrip strength ( Figure 3B , log-rank P<.001). Patients with low physical performance had a shorter OS than those with normal physical performance ( Figure 3C , log-rank P<.001). Moreover, patients with AWGS-defined sarcopenia had a shorter OS than those without AWGS-defined sarcopenia, whereas patients with AWGS-defined severe sarcopenia had the worst prognosis ( Figure 3D , log-rank P<.001).

Univariate and multivariable analyses with Cox proportional hazards models were performed to identify independent factors associated with OS ( Table 2 ). After adjustment for the same confounders, CT-defined sarcopenia (Model 1: HR, 2.00; 95% CI, 1.65-2.43) and AWGS-defined sarcopenia (Model 2: HR, 2.00; 95% CI, 1.59-2.49) were associated with poor prognosis. AWGS-defined severe sarcopenia indicated a higher risk of poor prognosis (Model 2: HR, 3.01; 95% CI, 2.21-4.09). All these indicators were more strongly associated with poor prognosis than low physical performance (PS score ≥2) (Model 3: HR, 1.37; 95% CI, 1.10-1.73).

The c statistics of the multivariable Cox Models were compared ( Table 2 ). The c statistics were 0.72 (95% CI, 0.69-0.74) for Model 1 and 0.76 (95% CI, 0.75-0.78) for Model 2, indicating moderate discrimination for OS. Model 2 was better than Model 1 in discriminating OS. The c statistic for Model 3 was 0.69 (95% CI, 0.67-0.72), indicating poor discrimination for OS.

To examine the robustness of our results, we performed an a priori sensitivity analysis by defining low SMM as the lowest quartile of the sex-specified study population and subsequently redetermined the proportions of CT-defined sarcopenia, AWGS-defined sarcopenia, and AWGS-defined severe sarcopenia accordingly. Afterwards, we reperformed univariate and multivariable analyses with Cox proportional hazards models ( eTable 2 in the Supplement ) and redrew the Kaplan–Meier curves ( eFigure 5 in the Supplement ). The results remained almost identical.

Lastly, we performed a priori subgroup analyses of multivariable Cox Models 1 and 2. Most of these analyses revealed that CT-defined sarcopenia, AWGS-defined sarcopenia, and AWGS-defined severe sarcopenia were all significantly associated with poor prognosis ( Figure 4 ). However, none of these conditions were significantly associated with poor prognosis in patients with large cell lung cancer.

Because chest CT scans are always available for patients with lung cancer, the clinical utility of chest CT scans, rather than abdominopelvic CT scans, is important in sarcopenia assessment. Furthermore, various organizations such as the US Preventive Services Task Force (32) and the Chinese Society of Clinical Oncology (33) have recommended low-dose chest CT scans for lung cancer screening in individuals with risk factors. The opportunistic utility of chest CT scans for screening lung cancer has been increasing to identify other diseases, such as chronic obstructive pulmonary disease and osteoporosis (34–36). Similarly, assessment of muscle health would be another opportunistic utility of these screening CT images.

The addition of handgrip strength and the PS score to chest CT-defined sarcopenia could further provide prognostic information on advanced NSCLC. Considering that the PS score is commonly obtained during routine oncological evaluation and that handgrip strength can be easily and repeatedly measured throughout cancer management without increasing any burden to patients, it is reasonable to use the combination of low SMM (derived from chest CT), handgrip weakness, and low physical performance (PS score ≥2) to define sarcopenia in the oncological research field.

Our study also has some limitations. First, our study was conducted at a single center; hence, the generalizability of our results seems to be limited. Second, in general, our study population might not be representative of patients with advanced NSCLC owing to potential referral bias. Third, we used a segmental multifrequency BIA device instead of dual-energy X-ray absorptiometry to estimate LBM. Nevertheless, according to a recent study, LBM measured by segmental multifrequency BIA has good agreement with dual-energy X-ray absorptiometry in ambulatory individuals (35). Fourth, we did not evaluate some important outcomes, including disease-specific mortality, quality of life, functional decline, and the incidence of adverse events related to chemotherapy.