Lung cancer is the most common cause of death from cancer worldwide. In 2013, the estimated global incidence of lung cancer was 1.8 million, and the estimated number of fatalities was 1.6 million [1]. Mortality from lung cancer has slightly decreased in Korea since 2002, but lung cancer remains the leading cause of death from cancer and a major health threat in Korea. The overall 5-year relative survival rate of lung cancer has improved from 11.2% for men and 14.7% for women among patients diagnosed between 1993 and 1997 to 19.3% for men and 28.2% for women among patients diagnosed between 2008 and 2012 [2,3]. The standard treatment for patients with stage I non–small-cell lung cancer (NSCLC) is anatomical surgical resection and appropriate mediastinal lymph node dissection; this treatment renders a chance of cure, with overall documented 5-year survival of up to 75% after surgery [4,5]. Perioperative management has improved over the past decades, but postoperative morbidity still occurs as the patient population ages and the complexity of the procedure increases alongside the more prevalent use of neoadjuvant therapies. Common respiratory complications after lobectomy include prolonged air leak, reported in 15% to 18% of cases, and pneumonia, reported in up to 6% [6].

Sarcopenia is a recently-defined geriatric syndrome that is characterized by an age-related decline in skeletal muscle mass combined with diminished muscle strength or reduced physical performance [7]. Sarcopenia is recognized as a predictor of poor surgical outcomes because it is associated with poor nutritional status and limited physical activity. Sarcopenia has been found to be associated with prolonged hospitalization and greater requirements of postoperative transfusion in elderly patients with femur fractures [8]. Studies in the field of thoracic surgery have also been performed, and while the effect of body mass index (BMI) as a predictor of surgical outcome remains unclear, Weig et al. [9] demonstrated that lean muscle mass was a more accurate predictor of postoperative outcome than BMI and that the lean psoas muscle area was a reliable predictor of surgical outcomes after lung transplantation. The psoas muscle area at the level of the third lumbar vertebra (L3) is commonly measured to assess sarcopenia [10,-12]. Assessment of the psoas muscle in patients with lung cancer is hindered since routine preoperative chest computed tomography (CT) usually does not include the lower lumbar area. Additionally, the area of the psoas muscle at L3 is not representative of the actual volume of the psoas muscle in every patient.

The purpose of this study was to assess the relationship between the area and the volume of the psoas muscle at L3 and to assess the feasibility of a psoas muscle volume index as a predictor of early postoperative complications after curative resection of early-stage NSCLC.

Patient cohort

Patients who underwent curative surgery for stage I/II NSCLC between January 2016 and December 2018 were retrospectively enrolled from our institution’s batabase (Seoul St. Mary’s Hospital, Seoul, Republic of Korea) of electronic medical records. We excluded patients who had undergone neoadjuvant therapy, who had pathologic stage III/IV cancer, who had undergone prior surgery for metachronous lung cancer, or for whom imaging studies were unavailable. The baseline clinical characteristics and preoperative bloodwork, including a carcinoembryonic antigen assay, as well as pulmonary function tests and imaging studies performed within 30 days prior to surgery were reviewed. Early postoperative complications, defined as morbidity occurring within 90 days of surgery, were analyzed together as overall complications. Additionally, respiratory complications, including prolonged (>5-day) air leak, recurrent pleural effusion, pneumonia, and bronchopleural fistula were analyzed.

This study was approved by the Institutional Review Board of Seoul St. Mary’s Hospital (approval no., KC19RESI0452). The requirement of informed consent from individual patients was waived because the study was a retrospective database review.

Image analysis

The psoas muscle area was measured at the region of L3 where the transverse process was most well-visualized on axial CT images taken from routine preoperative positron emission tomography-CT (PET-CT) scans. The psoas muscle volume was calculated semi-automatically from the level of the diaphragm to the symphysis pubis using image analysis software (Fujifilm SYNAPSE VINCENT; Fujifilm Medical, Tokyo, Japan). Subcutaneous and visceral fat tissues were identified, and the psoas muscle mass was calculated based on programmed Hounsfield units with an estimated threshold from –200 to –50 and exclusion of the fatty tissues (Fig. 1).

Since muscle mass can vary according to height and body weight, the calculated measurements must be normalized. Although there is no consistent method for the normalization of psoas muscle volume, we chose to use the cube of the height. In studies of human dynamics and physics, normalizing volume by the cube of height is the most widely accepted option of the various possible normalization methods [13]. Thus, to normalize the cross-sectional area, we defined the psoas area index (PAI) as the psoas muscle area at L3 divided by the square of height (m²), and to normalize the volume, we defined the psoas volume index (PVI) as the psoas muscle volume divided by the cube of height (m³).

Low-psoas volume index versus normal-to-high psoas volume index groups

Another key determinant of muscle mass is sex. We stratified the study cohort by sex for the analysis and identified low-PVI (PVI values in the lowest quartile) and normal-to-high–PVI (PVI values in the remaining quartiles) groups for each sex. We chose the lowest quartile as a separator because there is as yet no well-known threshold for defining sarcopenia by the psoas muscle volume.

Statistical analysis

The primary endpoint of this study was the relationship between PVI and all early (within 90 days of surgery) postoperative complications and between PVI and respiratory complications. As the secondary endpoint, we explored the relationship between PVI and preoperative pulmonary function, immunological status (neutrophil-to-lymphocyte ratio), and nutritional status via the Geriatric Nutritional Risk Index (GNRI). The GNRI was calculated using serum albumin concentration and body weight as 14.87×serum albumin concentration (g/dL)+41.7×(preoperative weight/ideal weight [kg]); ideal body weight was calculated as 22×the square of height (m²) [14,15]. The baseline characteristics were summarized using standard statistical methods, with data presented as mean±standard deviation for continuous variables and as frequencies with percentages for categorial variables. Distributions of continuous variables were compared using the Student t-test or the Mann- Whitney U-test depending on the result of the test of normality. Categorial variables were compared using the chi-square or Fisher exact test. The relationships between the PVI and PAI were assessed using Pearson correlation coefficients. The risk factors for overall complications and respiratory complications were analyzed using univariate and multivariate logistic regression. Statistical significance was indicated by a 2-sided p-value of <0.05, and variables determined to have statistical significance in the univariate analysis were incorporated into a multivariate model. All statistical analyses were performed with IBM SPSS Statistics for Windows ver. 24.0 (IBM Corp., Armonk, NY, USA), and the graphics were constructed using R ver. 3.4.1 (R Project for Statistical Computing, Vienna, Austria).

General characteristics of patients and psoas muscle profile

A total of 236 patients with pathologic stage I/II NSCLC who underwent curative pulmonary resection during the study period were eligible. Video-assisted thoracoscopic surgery was the predominant surgical approach (n=231, 97.9%). Common characteristics of the patients included adenocarcinoma (n=175, 74.2%), pathologic stage IA1-3 according to the 8th edition of tumor-node-metastasis staging system (n=151, 64.0%), pathologic stage N0 according to the same system (n=223, 94.5%), and lobectomy (n=204, 86.4%).

We assessed the relationship between psoas muscle area and volume based on the semi-automatic CT measurements of the psoas muscle area and volume at L3 (Fig. 2). A strong correlation between the 2 parameters (R²=0.816, p<0.001) was found and persisted after normalization (R²=0.743; p<0.001 for PAI and PVI).

Characteristics of the low- and normal-to-high–psoas volume index groups

The distribution of PVI by sex is shown in Fig. 3. We divided the cohort into 2 groups, according to the threshold separating the lowest quartile for sex-specific PVI (men, 71.37; women, 51.87) from the higher quartiles. The age at operation showed a weakly negative correlation with the PVI (R²=–0.368, p<0.001) (Fig. 4A). The 2 groups are compared in Table 1. The PAI (cm²/m²) was 3.96±1.14 in the low-PVI group and 5.44±1.63 in the normal-to-high–PVI group. The PVI (cm³/m³) was 52.86±11.08 in the low-PVI group and 79.92±19.02 in the normal-to-high–PVI group.

In the low-PVI group, the patients were older (p<0.001) and had a lower BMI (p=0.006), lower hemoglobin and albumin levels (p<0.001 for both), and poorer nutritional status, as reflected by a lower GNRI (p<0.001), than the patients in the normal-to-high group. Patients in the low-PVI group also had lower forced vital capacity (p=0.002), lower forced expiratory volume in 1 second (FEV1, p<0.001), and lower diffusing capacity for carbon monoxide (D_LCO, p<0.001) than those in the normal-to-high–PVI group (Fig. 4). There were no significant differences in the distribution of sex, history of tuberculosis or malignancy, type of operation, pathologic profile or stage of tumor, or length of postoperative hospital stay.

Early postoperative complications

The rate of all complications for the entire cohort was 39%, and the rate of respiratory complications was 27.1%. More complications were present in the low-PVI group (p=0.001) (Table 2), and postoperative atrial fibrillation (p=0.046) and vocal cord palsy (p=0.049) were significantly more frequent among patients with low PVI. Although statistically insignificant, prolonged air leaks, postoperative pneumonia, recurrent pleural effusion, and wound infection were also more frequently observed in the low-PVI group.

Risk factors for overall and respiratory complications

Univariate and multivariate logistic regression analyses were performed to identify potential risk factors for overall complications and for respiratory complications. On the univariate analysis for overall complications, male sex (OR, 2.078; p=0.007), low PVI (OR, 2.79; p=0.001), smoking history (OR, 1.019; p=0.002), and longer operation time (OR, 1.008; p=0.006) were associated with increased risk of complications, while high hemoglobin level (OR, 0.745; p=0.004), high albumin level (OR, 0.401; p=0.538), and high D_LCO (OR, 0.972; p=0.001) were associated with decreased risk (Table 3). Multivariate analysis found that low PVI (OR, 2.180; p=0.031) and preoperative hemoglobin level (OR, 0.686; p=0.002) were the strongest predictors of early postoperative complications.

We performed the same analyses focusing on only respiratory complications (Table 4). Univariate analysis showed an association between complication risk and male sex (OR, 2.593; p=0.001), low PVI (OR, 2.119; p=0.019), chronic obstructive pulmonary disease (OR, 3.796; p=0.008), smoking history (OR, 2.545; p=0.002), and longer operation time (OR, 1.010; p=0.001). High D_LCO (OR, 0.970; p= 0.002) was associated with a lower risk of respiratory morbidities. On multivariate analysis, only male sex (OR, 2.166; p=0.017) and a longer operation time (OR, 1.008; p=0.018) were associated with increased risk, and high D_LCO (OR, 0.974; p=0.008) remained associated with a decreased risk of respiratory morbidities in the early postoperative period.

Outcomes after lung cancer surgery and predictors thereof have been extensively reported [16], and several well-known predictors have been identified, namely age, male sex, pneumonectomy, low FEV1, and pre-existing comorbidities [17,-19]. With the changing profiles of patient populations, efforts to define preoperative risk factors have been increasingly focused on patient characteristics, mainly cardiac and respiratory reserve. As the population ages, much interest has been placed in assessing frailty and its relation to surgical outcomes. Frailty is characterized by decreased reserve and resistance to stressors resulting from cumulative declines across multiple physiological systems, causing vulnerability to adverse outcomes [20]. Decreased muscle mass is a result of frailty, and it is also a clinical marker in the assessment of frailty [21].

Among several candidates for predictors of surgical outcome, reduced muscle mass has been an interesting target, and its usefulness as a clinical marker for outcomes has been assessed in various settings. Sarcopenia is the progressive loss of skeletal muscle mass and strength associated with aging that has previously been shown to predict poor outcomes in patients who undergo abdominal, vascular, lung, and esophageal surgery [22,-25].

The levels and methods that most precisely reflect muscle mass have yet to be standardized. Since Mourtzakis et al. [26] showed that CT imaging of the muscle mass at L3 was as effective as dual-energy absorptiometry and superior to bioelectrical impedance analysis, the L3 muscle index has been used to assess sarcopenia. We also used L3 as the target level for measuring the psoas muscle area.

As far as we know, this is the first study to evaluate the relationship between area and volume in the psoas muscle and to explore the feasibility of the PVI as a predictor after lung cancer surgery. Volume can be more accurate for evaluating muscle mass than a single axial image, and we therefore used SYNAPSE 3D reconstruction software (Fujifilm Medical) to extract the entire psoas muscle volume in semi-automatic mode [27]. CT is ideal for identifying and quantifying skeletal muscle mass because it provides precise differentiation among fat, muscle, and other tissues. However, since routine preoperative chest CT in our lung cancer patients did not cover the lower lumbar area, we used CT images from routine preoperative PET-CT instead. It was possible to extract psoas muscle density from these images and to reconstruct and calculate the volume.

We found a strong correlation between psoas muscle area and volume, and we analyzed the risk factors for overall and respiratory complications. Because the muscle mass distribution is the most strongly affected by sex, we first normalized psoas area and volume (PAI, PVI) by height and then categorized the PVI by sex. As there is still no consensus regarding sex-specific cut-off values for sarcopenia, we arbitrarily defined a low PVI as the lowest quartile value.

Sarcopenia, as represented by a low PVI in this study, was found to be a negative prognostic factor for overall early postoperative complications. Sarcopenia is attributed to the aging process, but it has also been shown to be accelerated with malnutrition and chronic illness. The depletion of skeletal muscle mass leads to a decline in physical activity, which in turn leads to more profound sarcopenia. This vicious cycle leads to functional decline and increased susceptibility to comorbidities. Therefore, well-planned preoperative and rehabilitation programs that include physical exercise and nutritional support should help reduce complications in fragile patients. Other risk factors were found to include hemoglobin levels and smoking, which are well-known risk factors.

Interestingly, our study also showed that BMI was not a statistically significant risk factor. BMI does not account for the proportion of adipose tissue versus lean muscle mass, and it might not be an appropriate parameter to describe sarcopenia and frailty. Earlier research has shown that decreased BMI with preservation of muscle mass did not have a deleterious impact on survival [28].

Our results confirmed a significant correlation between sarcopenia and impaired pulmonary function. The association between low FEV1 and sarcopenia has been noted by others, as have the associations between low PVI and low FEV1, low forced vital capacity, and low D_LCO [9,27]. Pulmonary function is influenced by the interactions among the lungs, chest wall, and respiratory muscles. Among these, the respiratory muscle mass is positively correlated with the whole-body muscle mass, including the psoas muscle. Further research is warranted concerning the relationship between low skeletal muscle mass and D_LCO.

Another interesting result is that nutritional status was significantly correlated with sarcopenia. The GNRI, which is a function of serum albumin concentration and body weight, reflects nutritional status; a cross-sectional study of Chinese participants revealed that elderly patients with a low GNRI were more likely to have low muscle mass (OR, 3.904 for men and 4.486 for women; p<0.001 and p=0.001, respectively) [29]. These findings emphasize the need for nutritional support in addition to exercise training in sarcopenic patients to optimize their status.

This study had several limitations. First, it was a retrospective study conducted in a single institution. Second, it was a cross-sectional study based on PET-CT images that were taken at a single point in time before surgery. Trends or changes therefore could not be evaluated, and it is possible that patients with decreasing muscle mass would have worse outcomes. Third, because we used CT data from PET-CT images instead of contrast-enhanced abdominal CT, underestimation or overestimation of muscle mass was possible. However, the effect of this bias was minimized because we used the same imaging modalities for all patients. Therefore, the directionality of the conclusion should remain the same even if the magnitude of the effect changes slightly in further studies. Fourth and most importantly, because there was still no consensus definition of sarcopenia at the time of our study, we used the lowest quartile of the PVI.

In conclusion, the volume of the psoas muscle is a useful predictor of overall complication risk in the early postoperative period after curative resection in stage I/II NSCLC.

No potential conflict of interest relevant to this article was reported.

We extend special thanks to Sung Min Ko for his insightful comments regarding the methodology and the application of physics.

This study was supported by a Grant of the Samsung Vein Clinic Network (Daejeon, Anyang, Cheongju, Cheonan; Fund no. KTCS04-138).