Construction and evaluation of a prediction model for postoperative low muscle mass in older adult patients with gastric cancer

Abstract

Background

With global population aging, the prevalence of gastric cancer among patients aged 60 years and older continues to rise. Radical gastrectomy frequently induces postoperative skeletal muscle loss, which correlates with poor clinical prognosis. Currently, few prediction models are available to screen for new-onset postoperative low muscle mass in older gastric cancer patients. This study aimed to construct and internally validate a preoperative prediction model for low muscle mass at six months after radical gastrectomy in this elderly population.

Methods

A total of 69 eligible older adult patients with pTNM stage I–III gastric adenocarcinoma were retrospectively enrolled. Those with preoperative low muscle mass, neoadjuvant therapy, or incomplete clinical and imaging data were excluded. Postoperative low muscle mass was diagnosed using L3-level skeletal muscle index based on Asian cutoff criteria. Univariate and multivariate logistic regression analyses were applied to screen independent predictors. Model performance was assessed via ROC curves, calibration curves, and leave-one-out cross-validation (LOOCV).

Results

Among all participants, 17 patients (24.6%) developed postoperative low muscle mass. Preoperative BMI, skeletal muscle area, lymphocyte-monocyte ratio, and surgical method were confirmed as independent influencing factors. The established model yielded an AUC of 0.857, with a sensitivity of 94.1% and specificity of 67.3%, and the LOOCV AUC reached 0.832. Total gastrectomy was linked to a higher risk of low muscle mass compared with partial gastrectomy.

Conclusions

Preoperative BMI, SMA, and LMR can independently predict postoperative low muscle mass in older adult gastric cancer patients. The model shows moderate-to-good predictive performance in this single-center cohort and may help identify high-risk patients preoperatively. Prioritizing partial gastrectomy when oncologically safe and providing preoperative nutritional and exercise support may help reduce muscle loss.

Keywords

low muscle mass gastric cancer older adult gastrectomy prediction model inflammation skeletal muscle

1. Introduction

Gastric cancer remains the fifth most common malignant tumor and the third leading cause of cancer-related death worldwide, with a particularly high disease burden in the older adult population (≥60 years).¹ Radical gastrectomy is the standard curative treatment for resectable gastric cancer, but it disrupts gastrointestinal anatomy and physiology, reduces gastric reservoir function, impairs absorption of protein, vitamin B12, and other nutrients, and alters secretion of appetite-regulating hormones such as ghrelin.^2,3 These changes increase the risk of malnutrition and skeletal muscle loss in older adult patients.⁴

Postoperative low muscle mass affects 20%–40% of older adult patients with gastric cancer and is closely associated with anastomotic leakage, prolonged hospital stay, decreased quality of life, and reduced long-term survival.^5,6 Most existing studies focus on the prognostic impact of preoperative low muscle mass, while research on predicting new-onset postoperative low muscle mass remains insufficient.⁷ Current prediction models for gastric cancer-related outcomes rarely integrate CT-derived muscle mass parameters or inflammation-immune indicators, leaving an unmet need for risk stratification tools tailored to older adult patients.⁸

Early preoperative identification of high-risk patients allows timely interventions such as protein supplementation and resistance training to preserve muscle mass.^9,10 Surgical extent may also affect muscle loss: total gastrectomy leads to more severe nutrient malabsorption and stronger inflammatory responses than partial gastrectomy, which may accelerate muscle degradation.¹¹ However, the independent contribution of surgical method and its interaction with baseline characteristics remain unclear.

This study aimed to: 1) Identify preoperative clinical, anthropometric, and laboratory predictors of low muscle mass at 6 months after surgery in older adult patients with gastric cancer; 2) Develop and internally validate a logistic regression-based prediction model; 3) Explore the relationship between inflammation, muscle mass, and postoperative muscle loss; 4) Compare the risk of low muscle mass among different surgical approaches to support clinical decision-making.

2. Materials and methods

2.1. Study population

This retrospective study enrolled 69 older adult patients (≥60 years) who underwent radical gastrectomy for pathologically confirmed gastric adenocarcinoma (pTNM stages I–III) at The Second Affiliated Hospital of Soochow University between August 2017 and August 2024. The study was approved by the Ethics Committee of The Second Affiliated Hospital of Soochow University (No. 2021-068). The institutional ethics review board formally waived the requirement for written informed consent from all participants, as the study adopted a retrospective non-interventional design that only analyzed de-identified existing clinical and imaging data without direct contact with patients.

Inclusion Criteria: 1) Age ≥60 years; 2) American Society of Anesthesiologists (ASA) physical status classification < III¹²; 3) Pathologically confirmed gastric adenocarcinoma (pTNM stages I–III) with R0 resection; 4) Preoperative (within 1 week) and 6-month postoperative abdominal CT scans available for muscle measurement; 5) No preoperative low muscle mass (SMI > 40.8 cm²/m² for males, > 34.9 cm²/m² for females)¹³; 6) Complete clinical and laboratory data.

Exclusion Criteria: 1) Neoadjuvant chemotherapy, radiotherapy, or targeted therapy; 2) Metastatic, recurrent, or residual gastric cancer; 3) Previous gastrectomy or enterectomy; 4) Incomplete imaging or laboratory data.

2.2. Data collection

2.2.1. Baseline characteristics

Age, sex, smoking, drinking, comorbidities (hypertension, diabetes, chronic obstructive pulmonary disease [COPD]), pTNM stage (AJCC 8th edition),¹⁴ surgical method (total, proximal, distal gastrectomy), and postoperative hospital stay were collected.

2.2.2. Nutritional and laboratory indicators

Body mass index (BMI) was calculated as weight (kg)/height² (m²).¹⁵ The Controlling Nutritional Status (CONUT) score was used to evaluate nutritional status.¹⁶ Routine blood tests included albumin, prealbumin, hemoglobin, C-reactive protein (CRP), creatinine, lymphocytes, monocytes, etc. The lymphocyte-monocyte ratio (LMR) was calculated.

2.2.3. Muscle mass measurement

L3-level skeletal muscle area (SMA) and psoas major muscle area (PMMA) were measured on abdominal CT using ImageJ software. The skeletal muscle index (SMI) was calculated as SMA (cm²)/height² (m²).^17,18 Two experienced radiologists performed measurements independently; the intraclass correlation coefficient (ICC) was 0.92, indicating good consistency.

2.3. Statistical analysis

Data were analyzed using SPSS 21.0 and GraphPad Prism 9.0. Continuous variables were expressed as mean ± standard deviation or median (interquartile range). Categorical variables were expressed as case numbers and percentages. Univariate analysis was performed using t-test, Mann-Whitney U test, or χ² test. Variables with P<0.1 were entered into multivariate logistic regression. The model was evaluated using ROC curves, calibration curves, and LOOCV.^19,20 A two-sided P<0.05 was considered statistically significant.

3. Results

3.1. Baseline characteristics

A total of 69 older adult patients (≥60 years) were enrolled in this study, of whom 54 (78.3%) were male, with a median age of 71.0 years (interquartile range, 67.0–74.0 years). There were 24 (34.8%) patients undergoing total gastrectomy, 30 (43.5%) distal gastrectomy, and 15 (21.7%) proximal gastrectomy. pTNM stage distribution was 16 (23.2%) for stage I, 25 (36.2%) for stage II, and 28 (40.6%) for stage III. Baseline demographic, clinical, laboratory and nutritional characteristics between low muscle mass group and non-low muscle mass group are shown in Table 1. Baseline profiles were generally balanced except for variables entering the predictive model.

Table 1.

Baseline characteristics and univariate comparison between patients with and without postoperative low muscle mass.

Variable	Non-sarcopenia (n=52)	Sarcopenia (n=17)	Statistic	P-value
Gender, n (%)			χ² = 0.042	0.837
Female	11 (21.2)	4 (23.5)
Male	41 (78.8)	13 (76.5)
Age, years (mean ± SD)	71.87 ± 4.79	69.59 ± 5.30	t = 1.655	0.103
BMI, kg/m² (mean ± SD)	25.75 ± 3.14	22.58 ± 3.03	t = 3.646	0.001
SMA, mm² (mean ± SD)	12688.55 ± 2030.44	10841.53 ± 1823.69	t = 3.334	0.001
LMR (mean ± SD)	6.19 ± 1.38	4.88 ± 1.55	t = 3.318	0.004
CRP, mg/L (median [IQR])	2.50 [2.30–2.80]	5.40 [5.05–5.90]	Z = -5.248	<0.001
ALB, g/L (mean ± SD)	39.25 ± 3.18	37.86 ± 3.52	t = 1.542	0.128
Scr, μmol/L (median [IQR])	74.00 [63.00–80.00]	70.00 [58.00–78.00]	Z = -1.126	0.260
CONUT score (median [IQR])	2.00 [1.00–3.00]	3.00 [1.00–4.00]	Z = -1.528	0.126
pTNM staging, n (%)			Z = -2.064	0.039
I	19 (36.5)	1 (5.9)
II/III	33 (63.5)	16 (94.1)
Surgical method, n (%)			χ² = 9.562	0.008
Total gastrectomy	13 (25.0)	11 (64.7)
Proximal gastrectomy	14 (26.9)	1 (5.9)
Distal gastrectomy	25 (48.1)	5 (29.4)
Postoperative chemotherapy, n (%)	19 (36.5)	11 (64.7)	χ² = 4.136	0.042

Abbreviations: BMI = body mass index; SMA = skeletal muscle area; LMR = lymphocyte-monocyte ratio; CRP = C-reactive protein; ALB = albumin; Scr = creatinine; CONUT = Controlling Nutritional Status; pTNM = Pathological Tumor-Node-Metastasis.

3.2. Predictors of postoperative low muscle mass

Univariate analysis showed that BMI, SMA, LMR and surgical method were significantly associated with postoperative low muscle mass (all P<0.05). Variables with P<0.1 were included into binary logistic regression. Multivariate logistic regression identified preoperative BMI, SMA, LMR and surgical method as independent predictors, as summarized in Table 2. No obvious multicollinearity was observed among covariates. Compared with total gastrectomy, proximal and distal gastrectomy were independent protective factors.

Table 2.

Multivariate logistic regression analysis of independent predictors for postoperative low muscle mass.

Variable	β	SE	Wald χ²	OR	95% CI	P-value
BMI (kg/m²)	-0.392	0.171	5.242	0.676	0.483–0.945	0.022
SMA (mm²)	-0.001	0.000	5.046	0.999	0.999–1.000	0.025
LMR	-0.933	0.435	4.602	0.393	0.168–0.923	0.032
Surgical method			6.514			0.039
Total gastrectomy (ref)	0.000	0.000	0.000	1.000	-	-
Proximal gastrectomy	-2.857	1.414	4.083	0.057	0.004–0.918	0.043
Distal gastrectomy	-3.616	1.713	4.459	0.027	0.001–0.771	0.035
CRP (mg/L)	0.051	0.044	1.289	1.052	0.964–1.148	0.256
pTNM staging			2.562			0.278
Stage I (ref)	0.000	0.000	0.000	1.000	-	-
Stage II	0.566	1.545	0.134	1.761	0.085–36.344	0.714
Stage III	2.105	1.599	1.734	8.205	0.358–188.252	0.188
Postoperative chemotherapy	0.952	1.037	0.844	2.591	0.340–19.763	0.358

Abbreviations: β = regression coefficient; SE = standard error; OR = odds ratio; CI = confidence interval; ref = reference group.

3.3. Development and validation of the prediction model

Multivariate analysis identified three continuous independent predictors: BMI, SMA, and LMR. A simplified clinical prediction model was constructed: P = 0.356×BMI + 0.001×SMA + 0.599×LMR. To ensure reproducibility, the full logistic regression equation including intercept and dummy-coded surgical method is reported: Logit(P) = −15.236 + 0.356×BMI + 0.001×SMA + 0.599×LMR − 2.864×Proximal − 3.296×Distal. ROC curve analysis for the overall cohort is presented in Figure 1(a). ROC curve analysis for the overall cohort is presented in Figure 1(a). The model AUC was 0.857 (95%CI: 0.753–0.962, P<0.001), cut-off 24.91, sensitivity 94.1%, specificity 67.3%. The Hosmer–Lemeshow test indicated good calibration (χ²=5.236, P=0.682, Figure 2(a)). LOOCV AUC was 0.832.

Figure 1.

ROC curves of the prediction model (a) Overall population (n=69): AUC = 0.857 (95% CI: 0.753–0.962, P < 0.001). (b) Total gastrectomy subgroup (n=24): AUC = 0.797 (95% CI: 0.610–0.984, P = 0.014). (c) Distal gastrectomy subgroup (n=30): AUC = 0.984 (95% CI: 0.945–1.000, P = 0.001). The dashed line represents the reference (AUC = 0.5).

Figure 2.

Calibration curves of the prediction model (a) Overall population (n=69): Hosmer-Lemeshow test, χ² = 5.236, P = 0.682. (b) Total gastrectomy subgroup (n=24): Hosmer-Lemeshow test, χ² = 4.182, P = 0.524. (c) Distal gastrectomy subgroup (n=30): Hosmer-Lemeshow test, χ² = 1.836, P = 0.870. The solid line represents the ideal calibration (predicted probability = actual probability). The dashed line represents the model’s calibration.

Subgroup ROC results by surgical method are shown in Figure 1(b) and (c) and detailed predictive performance of the model and single predictors is listed in Table 3. Total gastrectomy subgroup AUC = 0.797; distal gastrectomy subgroup AUC = 0.984, which should be interpreted cautiously due to small sample size and limited events. The proximal gastrectomy subgroup had only one low muscle mass case, so ROC analysis was not performed. Calibration curves for subgroups are displayed in Figure 2(b) and (c).²¹

Table 3.

ROC curve analysis of the prediction model and individual predictors.

Population	Indicator	AUC	SE	P-value	95% CI	Optimal cut-off	Sensitivity (%)	Specificity (%)	Youden index
Overall (n=69)	BMI	0.782	0.070	0.001	0.644–0.919	21.72 kg/m²	58.8	90.4	0.492
	SMA	0.752	0.062	0.002	0.631–0.874	12689.82 mm²	82.4	57.7	0.401
	LMR	0.719	0.073	0.007	0.576–0.861	4.97	52.9	82.7	0.356
	Combined model	0.857	0.053	<0.001	0.753–0.962	24.91	94.1	67.3	0.614
Total gastrectomy (n=24)	Combined model	0.797	0.095	0.014	0.610–0.984	22.81	63.6	92.3	0.559
Distal gastrectomy (n=30)	Combined model	0.984	0.020	0.001	0.945–1.000	21.17	100.0	96.0

3.4. Postoperative changes in muscle and inflammatory markers

At 6 months postoperatively, all groups presented increased CRP, decreased creatinine and reduced LMR. The total gastrectomy group had the lowest psoas major muscle area (PMMA), indicating more severe central muscle loss. Postoperative PMMA differed significantly among three surgical approaches, while other muscle and inflammatory indices showed no significant intergroup differences.

4. Discussion

This study developed and internally validated a preoperative prediction model for 6-month postoperative low muscle mass in older adult patients with gastric cancer using three routinely available indicators: BMI, SMA, and LMR. The model showed moderate-to-good predictive performance in this single-center cohort. Total gastrectomy was an independent risk factor, supporting the choice of partial gastrectomy when oncologically feasible.

4.1. Key predictors

BMI reflects nutritional reserve; low BMI is associated with insufficient protein intake and reduced muscle synthesis.^22,23 SMA directly quantifies baseline muscle mass and is a reliable prognostic marker in gastric cancer.^24,25 LMR reflects immune-inflammatory status: high LMR indicates anti-inflammatory balance and reduced muscle breakdown.^26–28 Total gastrectomy causes more severe ghrelin deficiency, nutrient malabsorption, and inflammatory stress, leading to greater muscle loss.^3,29,30

4.2. Model application

The simplified formula enables rapid preoperative risk calculation. Patients with scores ≤24.91 may benefit from preoperative high-protein supplementation (1.2–1.5 g/kg/d) and resistance training.^31,32 Partial gastrectomy should be preferred when tumor location and stage permit.³³

4.3. Limitations

1. Small sample size and low statistical power: Only 69 patients and 17 events were included, which may reduce model stability and increase risk of overfitting. No formal sample size calculation was performed.

2. Single-center retrospective design: Selection bias and temporal heterogeneity in perioperative management cannot be excluded.

3. Outcome definition: Low muscle mass was defined by SMI only, not meeting EWGSOP2 criteria for sarcopenia, which require muscle strength or physical performance.⁴

4. Missing variables: Muscle strength, physical performance, dietary intake, and postoperative complications were not recorded.

5. Lack of external validation: The model requires validation in multi-center prospective cohorts.

6. Subgroup unreliability: The proximal gastrectomy subgroup had only one event, and the very high AUC in the distal gastrectomy subgroup should be interpreted cautiously.

5. Conclusions

Preoperative BMI, SMA, and LMR are independent predictors of postoperative low muscle mass in older adult patients with gastric cancer. The prediction model shows moderate-to-good performance in this single-center cohort and may help identify high-risk patients preoperatively. Total gastrectomy increases the risk of muscle loss. These findings are hypothesis-generating and require prospective validation before widespread clinical use.

Footnotes

ORCID iD

Shuhua Wu

Ethical considerations

This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of The Second Affiliated Hospital of Soochow University (No. 2021-068). The institutional ethics review board granted a formal waiver of written informed consent for all participants. The waiver was justified based on the retrospective, non-interventional nature of the research: all analyzed data were fully de-identified to eliminate patient identifiable information, and no additional clinical interventions, physical examinations, or direct patient contact were performed for this study.

Consent to participate

The institutional ethics review board granted a formal waiver of written informed consent for all participants. The waiver was justified based on the retrospective, non-interventional nature of the research: all analyzed data were fully de-identified to eliminate patient identifiable information, and no additional clinical interventions, physical examinations, or direct patient contact were performed for this study.

Author contributions

Linfang Li and Rong Zhang: Data curation, formal analysis, writing – original draft; Yuexuan Liu, Bai Li, Zhihui Zhang, Han Sun: Data collection, statistical analysis, visualization; Shuhua Wu: Conceptualization, methodology, supervision, writing – review and editing, funding acquisition.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Jiangsu Geriatric Health Research Project (LKZ2023007), Pre-research Fund of The Second Affiliated Hospital of Soochow University (SDFEYGZ2219), and Key Supported Medicine Discipline of Suzhou: Geriatrics (SZFCXK302113).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Bray

Ferlay

Soerjomataram

, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2018; 68(6): 394–424. https://doi.org/10.3322/caac.21492

Shim

Cheong

Lee

, et al. Perioperative nutritional status changes in gastrointestinal cancer patients. Yonsei Med J 2013; 54(6): 1370–1376. https://doi.org/10.3349/ymj.2013.54.6.1370

Kamiji

Troncon

Antunes-Rodrigues

, et al. Ghrelin and PYY(3-36) in gastrectomized patients: relations with appetite and energy intake. Eur J Clin Nutr 2010; 64(8): 845–852. https://doi.org/10.1038/ejcn.2010.88

Cruz-Jentoft

Bahat

Bauer

, et al. Sarcopenia: revised European consensus on definition and diagnosis. Age Ageing 2019; 48(1): 16–31. https://doi.org/10.1093/ageing/afz046

Kanazawa

Yamada

Kakinuma

, et al. Skeletal muscle mass depletion after gastrectomy negatively affects prognosis of gastric cancer patients. Anticancer Res 2020; 40(8): 4271–4279. https://doi.org/10.21873/anticanres.14429

Rinninella

Cintoni

Raoul

, et al. Muscle mass as a prognostic marker in gastric cancer: a systematic review and meta-analysis. Clin Nutr 2020; 39(7): 2045–2054.

Wagner

DeMarco

Amini

, et al. Frailty and sarcopenia in predicting outcomes after gastrointestinal surgery. World J Gastrointest Surg 2016; 8(1): 27–40. https://doi.org/10.4240/wjgs.v8.i1.27

Park

Kim

Seol

, et al. Prediction model for malnutrition risk after gastric cancer surgery. Ann Surg Oncol 2021; 28(8): 4471–4481. https://doi.org/10.1245/s10434-020-09559-3

Moisey

Mourtzakis

Cotton

, et al. Skeletal muscle predicts outcomes in older ICU patients. Crit Care 2013; 17(5): R206.

10.

Katsanos

Mazara

Apostolou

. Aerobic exercise for postprandial lipid metabolism. Sports Med 2006; 36(7): 547–560.

11.

Lee

Kim

Lee

, et al. Remnant stomach volume and nutrition after gastrectomy. Surg Oncol 2019; 31: 75–82.

12.

American Society of Anesthesiologists . ASA physical status classification system, 2024. Available from. https://www.asahq.org

13.

Yoon

Lee

Kim

, et al. Reference values for skeletal muscle index at L3 in healthy Koreans. Endocrinol Metab 2021; 36(3): 672–677.

14.

Amin

Edge

Greene

, et al. AJCC Cancer Staging Manual. 8th ed. Springer, 2017.

15.

Working Group on Obesity in China. Guidelines for prevention and control of overweight and obesity in Chinese adults. Chin J Epidemiol. 2004;25(1):1–6.

16.

Kondoh

Inoue

Yonezawa

, et al. Nutritional assessment using CONUT score in esophageal cancer. World J Surg 2018; 42(11): 3444–3451.

17.

Kim

, et al. Validation of automatic software for CT-derived skeletal muscle area. J Cachexia Sarcopenia Muscle 2018; 9(4): 664–672.

18.

Martin

Kentish-Barnes

Doherty

, et al. Body composition and outcome in critically ill patients. Intensive Care Med 2015; 41(3): 489–499.

19.

Hair

Black

Babin

, et al. Multivariate Data Analysis. 7th ed. Pearson, 2010.

20.

Arlot

Celisse

. A survey of cross-validation procedures for model selection. Stat Surv 2010; 4: 40–79. https://doi.org/10.1214/09-ss054

21.

Obuchowski

McClish

. Sample size determination for diagnostic accuracy studies. Stat Med 1997; 16(16): 1729–1742.

22.

Zhan

, et al. Postoperative skeletal muscle loss as a prognostic indicator in gastric cancer: a systematic review and meta-analysis. J Gastrointest Surg 2025; 29(2): 101898. https://doi.org/10.1016/j.gassur.2024.101898

23.

Curtis

Swan

Fox

, et al. Associations between BMI and probable sarcopenia in community-dwelling older adults. Nutrients 2023; 15(6): 1505. https://doi.org/10.3390/nu15061505

24.

Cruz-Jentoft

Landi

Schneider

, et al. Prevalence of and interventions for sarcopenia in ageing adults: a systematic review. Age Ageing 2014; 43(6): 748–759. https://doi.org/10.1093/ageing/afu115

25.

Borggreve

den Boer

van Boxel

, et al. Low muscle mass as a predictor of postoperative complications in gastric cancer: a systematic review and meta-analysis. J Clin Med 2020; 9(1): 199.

26.

Tomas

Eiriz

Vitorino

, et al. LMR as a prognostic biomarker in resectable gastric cancer. World J Gastrointest Oncol 2022; 14(7): 1307–1323.

27.

Svane

Boesen

Engel

. Cytotoxic T lymphocytes in tumor immune surveillance. Med Oncol 1999; 16(4): 223–238.

28.

Reid

. Cytokines and oxidative signalling in skeletal muscle. Acta Physiol Scand 2001; 171(3): 225–232. https://doi.org/10.1046/j.1365-201x.2001.00824.x

29.

Schmidt-Matthiesen

Oremek

. C-reactive protein in postoperative infectious complications. Chirurg 1990; 61(12): 895–899.

30.

Park

Kim

Lee

, et al. Surgical stress and inflammatory response after laparoscopic vs open gastrectomy: a systematic review and meta-analysis. Surg Endosc 2016; 30(10): 4323–4332.

31.

Japanese Gastric Cancer Association . Japanese gastric cancer treatment guidelines 2021 (5th edition). Gastric Cancer 2022; 25(1): 1–21.

32.

Peterson

Rhea

Sen

, et al. Resistance exercise for aging adults: a prescription for health. Am J Lifestyle Med 2010; 4(6): 509–523.

33.

Deutz

Bauer

Barazzoni

, et al. Protein intake and exercise for optimal muscle function with aging: a consensus statement. J Int Soc Sports Nutr 2014; 11: 38.