Skeletal muscle depletion during neoadjuvant chemoimmunotherapy predicts postoperative complications in non-small cell lung cancer: a retrospective cohort study

Abstract

Background

For patients with resectable locally advanced non-small cell lung cancer (NSCLC), surgical resection preceded by neoadjuvant chemoimmunotherapy (nICT) has become the standard treatment strategy. However, the triple physiological burden imposed by chemotherapy, immunotherapy, and ongoing tumor metabolism may compromise functional reserve and increase surgical risk. Identifying patients at high risk for postoperative complications remains a critical challenge in perioperative management. This study aimed to characterize dynamic changes in skeletal muscle and adipose tissue during nICT, to identify independent predictors of postoperative complications, and to evaluate the incremental predictive value of dynamic body composition metrics over established clinical variables.

Methods

This retrospective study included 160 patients with resectable NSCLC who underwent surgical resection following nICT between June 2024 and December 2025. Skeletal muscle index (SMI), skeletal muscle density (SMD), visceral adipose tissue (VAT), and subcutaneous adipose tissue (SAT) were quantified at the L3 vertebral level on computed tomography (CT) images obtained before and after treatment. Dynamic changes (Δ) were calculated as the difference between pre-treatment and post-treatment values. The primary outcome was postoperative complications of Clavien-Dindo (CD) grade ≥II. Three predictive models were constructed and compared using area under the receiver operating characteristic curve (AUC), net reclassification improvement (NRI), and integrated discrimination improvement (IDI).

Results

Following nICT, skeletal muscle metrics decreased significantly (median ΔSMI, −0.54 cm²/m²; P=0.01; median ΔSMD, −0.90 HU; P=0.003), while adipose tissue areas increased significantly (median ΔVAT, 9.54 cm²; P<0.001; median ΔSAT, 8.06 cm²; P<0.001). No significant correlation was observed between ΔSMI and ΔVAT (Spearman ρ=0.087; P=0.27), indicating divergent and independent metabolic trajectories. Of 160 patients, 58 (36.3%) developed CD grade ≥ II complications. Among the four dynamic body-composition measures, each evaluated in a separate multivariable model, only ΔSMI remained significantly associated with postoperative complications [adjusted odds ratio (OR) =0.806; 95% confidence interval (CI): 0.721–0.900; P<0.001], whereas ΔSMD, ΔVAT, and ΔSAT did not. Traditional assessments, including body mass index (BMI) change, Eastern Cooperative Oncology Group (ECOG) performance status, pulmonary function, and systemic inflammatory and nutritional indices, likewise showed no significant associations. Adding ΔSMI to the clinical model significantly improved AUC from 0.595 to 0.710 (ΔAUC =0.115; P=0.007).

Conclusions

Dynamic loss of skeletal muscle during nICT, as assessed on routine CT images, was independently associated with postoperative complications in patients with resectable NSCLC. Compared with baseline body-composition parameters and conventional clinical indices, longitudinal change in SMI (ΔSMI) provided better discrimination of perioperative risk. Because ΔSMI can be derived from CT scans already obtained for treatment-response assessment, it may serve as a practical, low-burden imaging biomarker to support preoperative risk stratification and identify patients who may benefit from targeted prehabilitation.

Highlight box.

Key findings

• Among four dynamic body-composition measures, only change in skeletal muscle index (ΔSMI) was independently associated with Clavien-Dindo grade ≥ II postoperative complications [adjusted odds ratio (OR) =0.806; 95% confidence interval (CI): 0.721–0.900; P<0.001].

• Neoadjuvant chemoimmunotherapy (nICT) was associated with divergent body-composition changes: skeletal muscle decreased while adipose tissue increased. Changes in muscle and visceral adipose tissue (ΔVAT) were uncorrelated (ΔSMI vs. ΔVAT, ρ=0.087, P=0.27).

• Adding ΔSMI to the clinical model improved area under the curve (AUC) from 0.595 to 0.710 (P=0.007); a combined muscle-adiposity phenotype further improved risk reclassification (NRI =0.418; IDI =0.049) without a significant additional AUC gain.

What is known and what is new?

• Preoperative sarcopenia is linked to worse postoperative and oncologic outcomes in non-small cell lung cancer (NSCLC). In gastrointestinal cancers, longitudinal body-composition changes during neoadjuvant therapy predict surgical risk beyond baseline measurements.

• This study characterizes body-composition changes during nICT in resectable NSCLC and identifies ΔSMI as a dynamic imaging biomarker independently associated with postoperative complications, with incremental predictive value beyond conventional clinical, nutritional, and inflammatory indices.

What is the implication, and what should change now?

• Static baseline metrics may not capture the decline in physiological reserve during neoadjuvant treatment; monitoring skeletal muscle change may improve perioperative risk assessment.

• ΔSMI can be assessed from routine treatment-response computed tomography scans without additional imaging. Patients with marked muscle loss may warrant closer assessment and prehabilitation (resistance exercise, nutritional optimization). Prospective studies should determine whether ΔSMI-guided interventions reduce postoperative complications.

Introduction

For patients with resectable locally advanced non-small cell lung cancer (NSCLC), surgery following neoadjuvant chemoimmunotherapy (nICT) has become an increasingly adopted treatment strategy. By reducing tumor burden and increasing the likelihood of complete resection, this multimodal approach has contributed to improved perioperative and oncologic outcomes. However, nICT also exposes patients to a period of considerable physiological stress. Cytotoxic chemotherapy, immune checkpoint blockade, treatment-related systemic inflammation, and the persistent metabolic demands of the tumor may collectively impair physiological reserve. In some patients, this decline may narrow the optimal window for surgery and increase susceptibility to postoperative complications or delayed recovery (1). Therefore, identifying patients at increased risk of postoperative morbidity remains an important challenge in perioperative decision-making.

Accurate assessment of physiological reserve is central to surgical risk stratification. In this context, computed tomography (CT)-based body composition analysis has emerged as an objective and reproducible approach. As recently reviewed (2,3), CT assessment at the third lumbar vertebral level is well suited for longitudinal evaluation of cancer-related body composition changes in patients with lung cancer, particularly because CT imaging is routinely incorporated into oncologic staging and response assessment. Unlike body mass index (BMI), which cannot distinguish lean tissue from adipose compartments, CT-based analysis allows compartment-specific quantification of skeletal muscle and fat distribution (4-6). In patients with NSCLC, preoperative sarcopenia has been associated with higher rates of postoperative complications and poorer long-term outcomes (7,8). Moreover, combining skeletal muscle quantity with muscle radiodensity, a surrogate marker of myosteatosis and intramuscular fat infiltration, may provide additional prognostic information for postoperative outcomes (6). Nevertheless, most available evidence has relied on static, single-timepoint measurements obtained before surgery or at baseline. Such assessments may not adequately capture the evolution of physiological reserve during neoadjuvant treatment. This limitation is particularly relevant in the nICT setting, in which treatment-related catabolism, inflammation, and changes in activity or nutrition may alter body composition over a relatively short interval. Notably, a recent study of patients with NSCLC undergoing nICT reported that static preoperative body composition metrics had limited prognostic value compared with baseline assessments (6). These findings suggest that the temporal change in body composition, rather than any single measurement alone, may contain clinically meaningful but underexplored information for perioperative risk assessment.

Recent evidence suggests that dynamic body composition changes during neoadjuvant therapy are superior to baseline static metrics in predicting postoperative outcomes for various solid tumors, including pancreatic (9), esophageal (10), and gastric cancers (11). Whether this prognostic relationship holds true for NSCLC patients receiving nICT, however, remains unknown. Given the distinct immunologic and metabolic effects of combined chemoimmunotherapy compared with chemotherapy alone, the specific pattern and clinical significance of treatment-induced body composition changes in this specific population warrant dedicated investigation, a gap explicitly noted in a recent narrative review synthesizing sarcopenia, systemic inflammation, and neoadjuvant therapy in resectable NSCLC (12).

The purpose of this study was threefold: first, to characterize dynamic changes in skeletal muscle and adipose tissue during nICT in patients with resectable NSCLC; second, to identify independent predictors of postoperative complications [Clavien-Dindo (CD) grade ≥II]; and third, to determine whether dynamic body composition metrics improve the predictive performance of clinical models for perioperative risk stratification beyond established clinical variables. We present this article in accordance with the STROBE reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1911/rc).

Methods

Study design and population

This retrospective cohort study consecutively included patients with NSCLC who underwent surgical resection following nICT. Patients were treated at the Lung Cancer Center of Sichuan University between June 2024 and December 2025. This study was approved by the Institutional Review Board of West China Hospital (Approval No. 2026-1506), and the requirement for informed consent was waived because this was a retrospective study. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Patients received programmed cell death 1 (PD-1) or programmed cell death ligand 1 (PD-L1) inhibitors combined with platinum-based doublet chemotherapy. Treatment response was evaluated after two cycles. Patients deemed suitable for continued treatment received one to two additional cycles, totaling three to four preoperative cycles. Surgical approaches included video-assisted thoracoscopic surgery (VATS) or thoracotomy.

Inclusion criteria were: (I) age 18–75 years; (II) histologically confirmed primary NSCLC; (III) receipt of at least two cycles of nICT; (IV) radical lung cancer resection; and (V) availability of pre-treatment and post-treatment contrast-enhanced chest and abdominal CT images. Exclusion criteria were: (I) a history of prior or concurrent other malignancies; (II) insufficient CT image quality for body composition segmentation; (III) incomplete clinical data; and (IV) prior thoracic radiotherapy.

Data collection and clinical variables

Clinical characteristics were extracted from electronic medical records. Demographics included age, sex, BMI, smoking status, Eastern Cooperative Oncology Group (ECOG) performance status, comorbidities, histologic subtype, and clinical stage. Treatment characteristics included the immunotherapy regimen, number of neoadjuvant treatment cycles, surgical approach, and surgical complexity. Complex resection was defined as pulmonary resection requiring any bronchoplastic procedure (including sleeve resection) and/or pulmonary arterial reconstruction.

Pretreatment laboratory data, including hemoglobin, serum albumin, white blood cell count, and absolute lymphocyte count, were extracted from electronic medical records within two weeks before initiation of nICT. The prognostic nutritional index (PNI) was calculated as: 10 × serum albumin (g/dL) + 0.005 × total lymphocyte count (/mm³)(13). The neutrophil-to-lymphocyte ratio (NLR) was calculated as the absolute neutrophil count divided by the absolute lymphocyte count. The systemic immune-inflammation index (SII) was calculated as platelet count × neutrophil count/lymphocyte count (14). Preoperative pulmonary function tests, including forced vital capacity (FVC), forced expiratory volume in one second (FEV₁), maximum voluntary ventilation (MVV), and diffusing capacity of the lung for carbon monoxide (DLCO), were performed within four weeks before surgery and expressed as percentages of predicted values.

Postoperative pathology was assessed using International Association for the Study of Lung Cancer (IASLC) criteria (15). Pathologic complete response (pCR) was defined as the absence of viable tumor cells in the resected specimen and sampled lymph nodes. Major pathologic response (MPR) was defined as 10% or less residual viable tumor cells in the primary tumor bed.

CT-based body composition analysis

Body composition was quantified using axial CT images at the third lumbar vertebral level (L3) (16). Pre-treatment assessment was performed within four weeks before treatment initiation, while preoperative assessment utilized the final CT scan before surgery. Image segmentation was conducted using SliceOmatic software (version 5.0; TomoVision, Montreal, Canada) by two trained radiologists blinded to clinical outcomes, with discrepancies resolved by consensus.

Tissue compartments were identified using standardized Hounsfield unit (HU) thresholds (17,18): skeletal muscle area (SMA: −29 to 150 HU), visceral adipose tissue (VAT: −150 to −50 HU), and subcutaneous adipose tissue (SAT: −190 to −50 HU). SMA included the psoas, erector spinae, quadratus lumborum, rectus abdominis, and abdominal oblique muscles. The skeletal muscle index (SMI) was calculated by normalizing SMA to height squared (cm²/m²). Skeletal muscle density (SMD) was defined as the mean HU attenuation of the SMA. Dynamic changes (△) were defined as the difference between post-treatment and pre-treatment values (e.g., △SMI = SMI_{post-treatment} – SMI_{pre-treatment}). Interobserver reliability was robust, with intraclass correlation coefficients (ICC) >0.99 for all measured variables.

Outcomes

The primary outcome was the occurrence of clinically relevant postoperative complications classified as CD grade ≥ II during a 3-month follow-up period (19). All 160 patients completed the 3-month follow-up period with no loss to follow-up. For patient-level analyses, patients with more than one postoperative complication were classified according to the highest CD grade observed. The operational definitions and grading rules for postoperative complications are provided in Table S1.

Statistical analysis

Normality was assessed using the Shapiro-Wilk test. Continuous variables were presented as mean ± standard deviation or median [interquartile range (IQR)], and categorical variables as frequencies (percentages). Intra-individual changes in body composition were compared using the Wilcoxon signed-rank test. Between-group comparisons were performed using the Mann-Whitney U test, χ² test, or Fisher’s exact test. Correlations were evaluated using Spearman’s rank coefficient. To identify independent predictors of complications, we performed univariable and multivariable logistic regression. ΔSMI, ΔSMD, ΔVAT, and ΔSAT were evaluated in four separate multivariable models, each including one body-composition measure together with the same set of predefined clinically relevant covariates: age, pretreatment BMI, clinical stage, smoking status, surgical approach, and FEV₁%. The combined muscle-adiposity phenotype and SMA-density phenotype were each entered as a categorical variable in separate multivariable models, adjusted for the same set of clinical covariates. Based on these results, three predictive models were constructed: Model 1 (clinical characteristics), Model 2 (Model 1 + △SMI), and Model 3 (Model 1 + muscle-adiposity phenotype). Internal validation was performed using 1,000 bootstrap resamples. Model discrimination was compared using the area under the receiver operating characteristic curve (AUC) and the DeLong test. Improvements in reclassification and discrimination were quantified via the continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI). Missing data were minimized by design: patients with incomplete clinical records or unavailable CT imaging were excluded prior to analysis as prespecified exclusion criteria. No imputation was performed, as no missing data were present among the 160 included patients. All analyses were performed using SAS Studio (version 3.81) and R (version 4.0.4), with two-sided P<0.05 considered statistically significant.

Results

Of 245 patients with NSCLC who underwent surgical resection following nICT, 85 were excluded for the following reasons: history of other malignancies (n=13), participation in other clinical trials (n=43), missing pre-treatment or post-treatment CT images (n=25), and failure of neoadjuvant targeted therapy (n=4). A total of 160 patients were included in the final analysis (Figure 1).

Figure 1.

Flowchart of patient selection. Of 245 patients with NSCLC who underwent surgical resection following neoadjuvant chemoimmunotherapy, 85 were excluded for the following reasons: history of other malignancies (n=13), participation in other clinical trials (n=43), missing pre-treatment or post-treatment CT images (n=25), and failure of neoadjuvant targeted therapy (n=4). A total of 160 patients were included in the final analysis. CT, computed tomography; NSCLC, non-small cell lung cancer.

No significant differences in pre-treatment characteristics were observed between patients with CD grade ≥II complications and those with gradeI complications (Table 1). The final cohort consisted of 152 men (95.0%) and 8 women (5.0%). Histologic types included squamous cell carcinoma in 138 patients (86.3%) and adenocarcinoma in 22 (13.7%). Initial clinical stages were stage IB (n=3, 1.9%), stage II (n=19, 11.9%), and stage III (n=138, 86.3%).

Table 1. Clinical and pathologic characteristics of patients stratified by postoperative complication status.

Pre-treatment characteristics	CD ≥II complications cohort (n=58)	CDcomplications cohort (n=102)	P value
Age, years	63.0±7.9	62.0±6.3	0.37
Gender			0.99
Male	56	97
Female	2	5
Smoking history	53	86	0.30
Pretreatment BMI, kg/m2	23.38±3.05	23.61±2.85	0.60
Preoperative BMI, kg/m2	23.71±3.30	23.92±3.00	0.63
ECOG PS			>0.99
0	53	93
1	5	9
≥2	0	0
Comorbidities
COPD	14	37	0.16
Coronary artery disease	9	23	0.31
Diabetes	6	16	0.48
Hypertension	20	41	0.50
Tumor type
Squamous cell carcinoma	49	89	0.89
Adenocarcinoma	9	13
TNM stage
IB/II	6	16	0.51
III	51	87
Pathological response			0.89
pCR	26	47
MPR	11	16
Non-MPR	24	36
Treatment cycles	3 [3–4]	3 [3–4]	0.94
Surgical approach			0.64
Minimally invasive	34	67
Open	21	38
Standard resection	43	80	0.67
Complex resection	15	22	0.56
Preoperative blood nutritional markers
Hemoglobin (g/L)	140.00 [128.00–147.00]	140.50 [128.00–150.00]	0.47
Albumin (g/L)	42.20 [39.40–44.60]	42.15 [39.88–44.90]	0.86
PNI	49.75 [45.75–53.10]	50.25 [47.15–53.10]	0.42
Preoperative blood inflammatory markers
WBC (×109/L)	6.90 [5.84–8.42]	7.37 [6.38–8.52]	0.23
NLR	2.98 [2.32–3.97]	3.02 [2.41–4.00]	0.88
SII	662.89 [464.18–940.57]	629.70 [445.72–1,069.30]	0.93
Preoperative pulmonary function
FVC (L)	3.49±0.59	3.32±0.59	0.32
FEV1(%)	88.79±20.05	88.42±19.79	0.95
DLCO (mL/min/mmHg)	20.90±4.06	21.59±4.70	0.59

Data are presented as n, mean ± standard deviation or median [interquartile range]. BMI, body mass index; CD, Clavien-Dindo; COPD, chronic obstructive pulmonary disease; DLCO, diffusing capacity of the lungs for carbon monoxide; ECOG PS, Eastern Cooperative Oncology Group performance status; FEV₁, forced expiratory volume in 1 second; FVC, forced vital capacity; MPR, major pathologic response; NLR, neutrophil-to-lymphocyte ratio; pCR, pathologic complete response; PNI, prognostic nutritional index; SII, systemic immune-inflammation index; TNM, tumor-node-metastasis; WBC, white blood cell count.

The mean pretreatment BMI was 23.53±2.91 kg/m², and the preoperative BMI was 23.84±3.10 kg/m². A smoking history was present in 139 patients (86.9%). All patients had an ECOG performance status of 0 or 1. Comorbidities included chronic obstructive pulmonary disease (n=51, 31.9%), coronary artery disease (n=32, 20.0%), diabetes (n=22, 13.8%), and hypertension (n=61, 38.1%).

The median number of chemoimmunotherapy cycles was 3 (IQR: 3–4). Specifically, 78 patients (48.8%) received three cycles and 82 (51.2%) received four. Regarding surgical approach, 99 patients (61.9%) underwent minimally invasive surgery and 61 (38.1%) underwent thoracotomy. A total of 37 patients (23.1%) underwent complex resection, defined as resection requiring bronchoplastic procedures, including sleeve resection and/or pulmonary arterial reconstruction. pCR was achieved in 73 patients (45.6%), and MPR was achieved in 27 (16.9%).

For lung function, there was no significant difference in FVC (3.49±0.59 vs. 3.32±0.59 L, P=0.32), FEV₁% (88.79%±20.05% vs. 88.42%±19.79%, P=0.95) and DLCO (20.90±4.06 vs. 21.59±4.70 mL/min/mmHg, P=0.59).

Similarly, no significant differences were observed in pretreatment blood markers (all P>0.05). Nutritional indicators, such as hemoglobin [140.00 (128.00–147.00) vs. 140.50 (128.00–150.00) g/L, P=0.47], albumin [42.20 (39.40–44.60) vs. 42.15 (39.88–44.90) g/L, P=0.86], and PNI [49.75 (45.75–53.10) vs. 50.25 (47.15–53.10), P=0.42], were comparable between the two groups. Furthermore, systemic inflammatory markers, including white blood cell (WBC) count [6.90 (5.84–8.42) ×10⁹/L vs. 7.37 (6.38–8.52) ×10⁹/L, P=0.23], NLR [2.98 (2.32–3.97) vs. 3.02 (2.41–4.00), P=0.88], and SII [662.89 (464.18–940.57) vs. 629.70 (445.72–1,069.30), P=0.93], showed no statistical association with the severity of complications.

Following nICT, body composition parameters demonstrated significant dynamic changes (Figure 2, Table 2). Mean BMI increased from 23.53±2.91 kg/m² at pre-treatment to 23.84±3.10 kg/m² at post-treatment (mean increase, 0.32 kg/m²). Both skeletal muscle metrics significantly decreased during treatment: the median SMI decreased from 48.86 cm²/m² (IQR: 43.14–55.08 cm²/m²) to 47.71 cm²/m² (IQR: 41.91–53.91 cm²/m²) (P=0.01), and the median SMD decreased from 40.19 HU (IQR: 36.80–43.28 HU) to 39.20 HU (IQR: 35.88–42.70 HU) (P=0.003).

Figure 2.

CT-based body composition segmentation at the L3 level before and after neoadjuvant chemoimmunotherapy. Representative images from a 59-year-old man with resectable NSCLC who completed four treatment cycles. Tissue compartments are color-coded as follows: red, skeletal muscle; yellow, VAT; green, SAT. (A) Baseline axial CT image. (B) Preoperative axial CT image. (C) Baseline segmentation overlay (SMI, 49.09 cm²/m²; SMD, 35.92 HU; VAT, 303.1 cm²; SAT, 149.2 cm²). (D) Preoperative segmentation overlay (SMI, 40.79 cm²/m²; SMD, 30.74 HU; VAT, 312.5 cm²; SAT, 142.8 cm²). CT, computed tomography; HU, Hounsfield unit; NSCLC, non-small cell lung cancer; SAT, subcutaneous adipose tissue; SMD, skeletal muscle density; SMI, skeletal muscle index; VAT, visceral adipose tissue.

Table 2. Dynamic changes in body composition parameters before and after neoadjuvant chemoimmunotherapy in the overall cohort (n=160).

Body composition parameters	Pre-treatment	Post-treatment	Absolute change	Relative change percent, %	P value
BMI, kg/m2	23.53±2.91	23.84±3.10	0.32±1.56	1.49±6.72	0.007
SMI, cm2/m2	48.86 (43.14–55.08)	47.71 (41.91–53.91)	−0.54 (−2.52 to 1.49)	−1.14 (−5.55 to 3.20)	0.01
SMD, HU	40.19 (36.80–43.28)	39.20 (35.88–42.70)	−0.90 (−3.06 to 1.10)	−2.37 (−7.08 to 2.59)	0.003
VAT, cm2	111.80 (66.36–167.65)	122.50 (79.17–182.90)	9.54 (−0.59 to 27.30)	7.77 (−0.31 to 23.36)	<0.001
SAT, cm2	101.60 (82.56–130.95)	114.40 (85.88–136.75)	8.06 (−2.43 to 17.70)	7.43 (−2.11 to 21.68)	<0.001

Data are presented as mean ± standard deviation or median (interquartile range). BMI, body mass index; HU, Hounsfield unit; SAT, subcutaneous adipose tissue; SMD, skeletal muscle density; SMI, skeletal muscle index; VAT, visceral adipose tissue.

Conversely, both adipose tissue metrics significantly increased. The median VAT area increased from 111.80 cm² (IQR: 66.36–167.65 cm²) to 122.50 cm² (IQR: 79.17–182.90 cm²) (P<0.001), and the median SAT area increased from 101.60 cm² (IQR: 82.56–130.95 cm²) to 114.40 cm² (IQR: 85.88–136.75 cm²) (P<0.001). No significant correlation was observed between ΔSMI and ΔVAT (Spearman ρ=0.087; P=0.27).

Among the 160 patients, 58 (36.3%) developed postoperative complications of CD grade ≥II within 90 days after surgery. According to the highest CD grade observed per patient, 48 (30.0%) had grade II complications, 7 (4.4%) had grade IIIa complications, and 3 (1.9%) had grade IIIb complications; no grade IV or V complications occurred (Tables S2,S3).

Univariable analysis indicated that the ΔSMI was significantly lower in the complication group (CD ≥II) than in the group without major complications [−1.90 vs. −0.02 cm²/m²; P<0.001] (Table S4). A significant association was observed between ΔSMI and postoperative complications [odds ratio (OR) =0.845; 95% confidence interval (CI): 0.769–0.928; P<0.001]. Conversely, ΔSMD, ΔVAT, and ΔSAT showed no significant associations (all P>0.05).

In the multivariable logistic regression analysis, adjusted for age, pretreatment BMI, clinical stage, smoking status, surgical approach, and FEV₁%, ΔSMI remained an independent risk factor for CD ≥II complications (OR =0.806; 95% CI: 0.721–0.900; P<0.001) (Table 3 and Table S5).

Table 3. Association between dynamic changes in body composition parameters and postoperative complications of CD Grade ≥II.

Body composition parameters	Non/CD =I (n=102)	CD ≥II (n=58)	OR (95%CI)	Adjusted OR (95% CI)	P value
ΔSMI, cm2/m2	−0.02 (−1.70 to 2.30)	−1.90 (−5.06 to −0.03)	0.845 (0.769–0.928)	0.806 (0.721–0.900)	<0.001
ΔSMD, HU	−0.80 (−2.50 to 1.29)	−1.60 (−3.33 to −0.75	0.974 (0.900–1.060)	1.017 (0.961–1.070)	0.58
ΔVAT, cm2	8.60 (−0.97 to 28.12)	11.65 (1.89 to 25.87)	1.003 (0.990–1.017)	1.004 (0.987–1.021)	0.66
ΔSAT, cm2	8.38 (−1.73 to 17.94)	5.05 (−3.58 to 17.27)	1.002 (0.986–1.018)	1.018 (0.995–1.041)	0.79

CD, Clavien-Dindo; HU, Hounsfield unit; SAT, subcutaneous adipose tissue; SMD, skeletal muscle density; SMI, skeletal muscle index; VAT, visceral adipose tissue.

For the CD grade III or higher endpoint, ΔSMI was associated with increased odds of major complications in univariable analysis (OR, 0.776; 95% CI: 0.653–0.923; P=0.004) and in Firth penalized multivariable logistic regression (adjusted OR, 0.788; 95% CI: 0.658–0.943; P=0.009) (Table 4).

Table 4. Grade-specific associations between ΔSMI and postoperative complications.

Outcome	Number of events	OR (95% CI)	Adjusted OR (95%CI)	P value
CD ≥II vs. CD	58	0.845 (0.769–0.928)	0.806 (0.721–0.900)	<0.001
CD II vs. CD	48	0.838 (0.760–0.924)	0.846 (0.759–0.942)	0.002
CD ≥III vs. CD	10	0.766 (0.653–0.923)	0.788 (0.658–0.943)	0.009

CD, Clavien-Dindo; CI, confidence interval; OR, odds ratio; SMI, skeletal muscle index.

To determine if adipose tissue changes modified the effect of muscle depletion, patients were categorized into four phenotypes based on the direction of ΔSMI and ΔVAT changes (Table 5 and Table S6). After multivariable adjustment, the group with decreased muscle and increased fat had the highest complication risk compared with the reference group (stable/increased muscle and stable/increased fat) (OR =5.75; 95% CI: 2.42–13.65; P<0.001). The group with concurrent decreases in both muscle and fat also exhibited an increased risk (OR =3.41; 95% CI: 1.12–10.36; P=0.03).

Table 5. Postoperative complication rates across combined muscle-adiposity phenotypes.

Phenotype	N	CD ≥II	OR (95% CI)	Adjusted OR (95% CI)	P value
ΔSMI ≥0/ΔVAT ≥0	60	16 (26.7)	1	1
ΔSMI <0/ΔVAT ≥0	63	30 (47.6)	2.73 (1.30–5.74)	5.75 (2.42–13.65)	<0.001
ΔSMI ≥0/ΔVAT <0	6	1 (16.7)	0.46 (0.20–1.88)	0.69 (0.19–2.50)	0.57
ΔSMI <0/ΔVAT <0	31	11 (35.5)	3.44 (1.59–7.47)	3.41 (1.12–10.36)	0.03

Data are presented as n (%) unless otherwise indicated. CD, Clavien-Dindo; CI, confidence interval; OR, odds ratio; SMI, skeletal muscle index; VAT, visceral adipose tissue.

Formal interaction analysis demonstrated that an increase in ΔVAT did not significantly modify the association between ΔSMI decrease and postoperative complications (P for interaction =0.21), indicating that the effect of ΔSMI on complication risk was statistically independent of concurrent ΔVAT changes. A comparison between the ΔSMI <0/ΔVAT ≥0 group and the ΔSMI <0/ΔVAT <0 group showed no significant difference (relative OR =1.685; P=0.35).

A combined SMA-density phenotype (ΔSMI + ΔSMD) was further constructed (Table 6 and Table S7). Multivariable regression analysis demonstrated that only the concurrent ΔSMI and ΔSMD decrease group showed a statistically significant increase in complication risk compared with the reference group (ΔSMI ≥0/ΔSMD ≥0) (adjusted OR =2.46; 95% CI: 1.06–5.72; P=0.04), whereas the isolated ΔSMI decrease group did not reach statistical significance (adjusted OR =1.93; 95% CI: 0.60–6.14; P=0.27). An isolated decrease in ΔSMD showed no predictive value (adjusted OR =0.55; 95% CI: 0.15–2.01; P=0.37).

Table 6. Postoperative complication rates across combined skeletal muscle area-density phenotypes.

Phenotype	N	CD ≥II	OR (95% CI)	Adjusted OR (95% CI)	P value
ΔSMI ≥0/ΔSMD ≥0	42	11 (26.2)	1	1
ΔSMI <0/ΔSMD ≥0	19	8 (42.1)	2.05 (0.65–6.42)	1.93 (0.60–6.14)	0.27
ΔSMI ≥0/ΔSMD <0	24	4 (16.7)	0.56 (0.16–2.02)	0.55 (0.15–2.01)	0.37
ΔSMI <0/ΔSMD <0	75	35 (46.7)	2.47 (1.08–5.62)	2.46 (1.06–5.72)	0.04

Data are presented as n (%) unless otherwise indicated. CD, Clavien-Dindo; CI, confidence interval; OR, odds ratio; SMD, skeletal muscle density; SMI, skeletal muscle index.

Three predictive models were constructed: Model 1 (clinical model, including age, pretreatment BMI, clinical stage, smoking history, surgical approach, and FEV₁%) with an AUC of 0.595; Model 2 (clinical model + ΔSMI) with an AUC of 0.710; and Model 3 (clinical model + combined muscle-adiposity phenotype) with an AUC of 0.743 (Figure 3).

Figure 3.

Predictive performance and clinical utility of three models for postoperative complication risk stratification. (A) ROC curves for the three predictive models for postoperative complications of Clavien-Dindo grade ≥II. (B) DCA for the three predictive models for postoperative complications of Clavien-Dindo grade ≥II. AUC, area under the receiver operating characteristic curve; DCA, decision curve analysis; ROC, receiver operating characteristic; SMI, skeletal muscle index.

Decision curve analysis (DCA) demonstrated that across a threshold probability range of 10% to 60%, the net benefits of both Model 2 and Model 3 were higher than that of Model 1.

Following internal validation with 1,000 bootstrap resamples, Model 2 demonstrated significant improvements over Model 1 (ΔAUC =0.115, P=0.007; continuous NRI =0.4228, P<0.001). Similarly, Model 3 significantly outperformed Model 1 (ΔAUC =0.148, P=0.002; NRI =0.7484, P<0.001).

Compared with Model 2, Model 3 showed no statistically significant difference in AUC (ΔAUC =0.033, P=0.24). However, the improvements in both NRI (0.418, P=0.01) and IDI (0.049, P=0.02) were statistically significant, indicating that the combined phenotype can further improve risk reclassification capability.

Discussion

Principal findings and predictive superiority of ΔSMI

This study characterized longitudinal changes in body composition during nICT in patients with resectable NSCLC and identified the dynamic change in SMI (ΔSMI) as an independent predictor of clinically relevant postoperative complications, defined as CD grade II or higher (adjusted OR, 0.806; 95% CI: 0.721–0.900; P<0.001), after adjustment for age, pretreatment BMI, clinical stage, smoking history, surgical approach, and FEV₁% predicted. Importantly, this association was not driven solely by lower-severity events: in an exploratory sensitivity analysis restricted to major complications of CD grade III or higher, lower ΔSMI remained significantly associated with increased risk (adjusted OR, 0.788; 95% CI: 0.658–0.943; P=0.009). This consistency across the severity spectrum supports the clinical relevance of ΔSMI beyond the primary CD grade II or higher endpoint.

Notably, conventional functional and clinical indicators, including BMI change and ECOG performance status, were not significantly associated with the study-defined endpoint in this cohort. This finding should be interpreted cautiously, given that the primary endpoint encompassed a range of clinical severity from grade II to grade IIIb events. Nevertheless, the observation that ΔSMI remained associated with both the composite endpoint and the more restrictive major-complication endpoint suggests that CT-derived longitudinal body composition assessment may capture treatment-related deterioration in physiological reserve that is not readily reflected by these conventional clinical metrics.

Dynamic body composition trajectories vs. static snapshots

Prior studies of body composition in NSCLC have largely focused on static, single-timepoint preoperative assessments. Lee et al. (5) demonstrated in a cohort of 7,076 patients that combined evaluation of skeletal muscle mass and radiodensity at a single preoperative timepoint had significant prognostic value for long-term postoperative survival. More recently, Zheng et al. assessed baseline and preoperative CT-derived body composition in 154 patients with NSCLC receiving nICT, a population closely comparable to ours. They reported that baseline body composition parameters were independently associated with disease-free survival, whereas static preoperative metrics showed limited prognostic value (20).

Our findings provide a complementary perspective on these observations. When body composition is considered as a longitudinal trajectory rather than a static measurement, the extent of skeletal muscle change during treatment—represented by ΔSMI—appears to carry clinically relevant information for predicting postoperative complications. This result is consistent with studies in esophageal (10), gastric (11) and NSCLC (21) cohorts, in which posttreatment or treatment-related body composition changes have shown stronger associations with surgical outcomes than baseline measurements alone.

An important distinction between our findings and those of Zheng et al. concerns the relative contribution of muscle quantity and muscle quality. Zheng et al. identified muscle density, expressed as pectoralis muscle density (PMD), and adipose tissue parameters as key prognostic features, whereas skeletal muscle quantity did not show independent prognostic value in their cohort (20). In contrast, our results suggest that, for short-term postoperative morbidity, acute loss of skeletal muscle mass, rather than change in muscle radiodensity, may be more closely associated with surgical risk. This discrepancy may reflect differences in clinical endpoints—long-term survival versus immediate postoperative complications—as well as differences in the biological timescales over which muscle mass and muscle quality respond to treatment-related catabolic stress.

Mechanisms underlying divergent muscle and adipose trajectories

The concurrent skeletal muscle depletion and adipose tissue accumulation observed in our cohort reflect divergent body composition trajectories driven by largely independent mechanisms. Skeletal muscle depletion during chemoimmunotherapy results from the convergence of multiple catabolic signals: platinum-based chemotherapy activates the ubiquitin–proteasome system and autophagy pathways, suppressing muscle protein synthesis while accelerating myofibrillar degradation (22); concurrently, programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) inhibitor-mediated immune activation amplifies systemic inflammatory cytokine cascades, particularly tumor necrosis factor-alpha (TNF-α) and interleukin-6 (IL-6), which further upregulate NF-κB and JAK-STAT signaling, accelerating myofibrillar catabolism and impairing muscle regenerative capacity (23). The concurrent increase in adipose tissue likely reflects chemotherapy-induced suppression of brown adipose tissue thermogenic activity, which reduces total energy expenditure and promotes lipid accumulation in visceral and subcutaneous depots (24). The absence of a significant correlation between ΔSMI and ΔVAT (Spearman ρ=0.087; P=0.27) provides empirical support for the independence of these two metabolic trajectories, consistent with the pattern observed in patients with locally advanced rectal cancer undergoing neoadjuvant chemoradiotherapy (25), and distinct from the adipose tissue loss with modest muscle gain reported in pancreatic cancer (9) and the muscle depletion without adipose divergence seen in ovarian and esophageal cancers (26), highlighting that the direction and magnitude of treatment-induced body composition changes vary substantially by tumor type and treatment regimen.

Clinical implications and incremental predictive value of dynamic skeletal muscle assessment

Treatment-related skeletal muscle depletion reflects more than isolated muscle wasting; it may indicate a broader disruption of the muscle-immune-metabolic axis governing surgical recovery (27), and may compromise respiratory muscle function, cough effectiveness, airway clearance, and postoperative mobilization, thereby increasing susceptibility to pulmonary complications and delayed recovery. Although these mechanisms could not be directly evaluated in this retrospective study, the observed association suggests that longitudinal muscle loss during nICT may provide clinically relevant information beyond a single preoperative assessment.

From a clinical implementation perspective, ΔSMI can be derived from CT scans already obtained for treatment-response assessment and therefore does not require additional imaging, radiation exposure, or substantial patient burden. In our cohort, adding ΔSMI to a clinical model containing age, pretreatment BMI, clinical stage, smoking history, surgical approach, and FEV₁% predicted improved discrimination for clinically relevant postoperative complications, with the AUC increasing from 0.595 to 0.710 (ΔAUC =0.115; P=0.007). The consistency of this association across both the composite endpoint and the more restrictive major-complication endpoint suggests that ΔSMI may help identify patients at risk of clinically significant, rather than solely minor, postoperative morbidity. These findings suggest that CT-derived skeletal muscle change may complement, rather than replace, established clinical assessment in perioperative risk stratification. In particular, patients with marked skeletal muscle depletion during neoadjuvant treatment may benefit from closer preoperative assessment and consideration of individualized nutritional support and exercise-based prehabilitation. However, given the retrospective single-center design and the limited sample size, ΔSMI should not be used in isolation to determine surgical eligibility. Prospective multicenter studies are needed to validate its predictive performance and to determine whether ΔSMI-guided interventions can reduce postoperative morbidity.

Beyond ΔSMI alone, combining muscle and adiposity or muscle-quality metrics provided modest additional discriminatory information. Patients with concurrent muscle depletion and visceral fat accumulation (ΔSMI <0/ΔVAT ≥0) had the highest complication risk relative to the reference group with stable or increased muscle and fat (adjusted OR, 5.75; 95% CI: 2.42–13.65). Incorporating this muscle-adiposity phenotype into the clinical model (Model 3) improved risk reclassification beyond ΔSMI alone (AUC =0.743, NRI =0.418; P=0.01; IDI =0.049; P=0.02), although the increase in AUC over the ΔSMI-based model was not statistically significant (ΔAUC =0.033; P=0.24). Similarly, concurrent decline in both muscle quantity and muscle density (ΔSMI <0/ΔSMD <0) was associated with increased complication risk (adjusted OR, 2.46; 95% CI: 1.06–5.72), whereas an isolated decrease in either measure alone did not reach statistical significance. Taken together, these exploratory findings suggest that composite muscle-adiposity phenotyping may offer incremental value for risk reclassification beyond a single continuous measure, but larger prospective studies are needed to confirm their clinical utility and to determine whether the added complexity of phenotype-based stratification justifies its use over ΔSMI alone.

Clinical implications: transitioning toward a host-centric phenotype

For patients with marked skeletal muscle loss during neoadjuvant treatment, clinicians may consider reassessing surgical tolerance and initiating targeted prehabilitation, including nutritional optimization and resistance exercise training (28-30), before surgery. This approach is consistent with the framework proposed by Verkoulen et al. (3), which emphasizes integration of CT-derived body composition metrics with host-level physiological assessment. Such a strategy may help move perioperative evaluation beyond tumor-centric staging toward a host-centered phenotype that better reflects individual surgical risk.

Our findings support the potential role of dynamic body composition trajectories, particularly ΔSMI, as a measurable component of this host phenotype in patients undergoing nICT. Because ΔSMI can be obtained from routine response-assessment CT scans, it may provide a practical imaging marker to inform perioperative risk assessment and guide future interventional studies.

Study limitations and future research directions

This study has several limitations. First, its retrospective, single-center design may limit generalizability; multicenter prospective validation is needed. Second, the cohort was predominantly male (95.0%) and composed mainly of patients with squamous cell carcinoma (86.3%), so the applicability of our findings to female patients and those with adenocarcinoma requires further evaluation. Third, body composition was assessed using a single axial CT slice at the L3 vertebral level, which may not fully represent regional or whole-body variation in muscle and adipose tissue distribution. Fourth, the primary endpoint spanned CD grade II through IIIb, and the exploratory analysis restricted to grade III or higher events included only 10 patients; this estimate warrants confirmation in larger cohorts.

Conclusions

In conclusion, dynamic skeletal muscle depletion during nICT quantified via routine CT imaging is an independent risk factor for postoperative complications in patients with NSCLC, with predictive efficacy significantly exceeding that of static baseline metrics and traditional clinical functional scores. ΔSMI assessment can be seamlessly integrated with routine CT evaluations, rendering it a practical tool for perioperative risk stratification. Future prospective studies should evaluate whether ΔSMI-guided prehabilitation strategies can meaningfully improve surgical outcomes in this high-risk population.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of West China Hospital (Approval No. 2026-1506), and the requirement for informed consent was waived because this was a retrospective study.

Data Sharing Statement

Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1911/dss

jtd-18-9-1018-dss.pdf ^{(86.1KB, pdf)}

DOI: 10.21037/jtd-2026-1911