Prognostic Value of Early Weight Gain in Patients With Stage IV Gastric Cancer Undergoing First-Line Combination Immunotherapy

Abstract

Purpose

Conventional radiologic assessment using Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1) may not fully capture clinical responses during immunotherapy. Therefore, we investigated the association between early weight gain (WG) and prognosis in this population.

Materials and Methods

We enrolled 268 consecutive patients with stage IV gastric cancer (GC) who received first-line combination immunotherapy. Data were obtained from a prospective clinical database. Cox regression and g-computation were used to quantify and visualize the associations between WG and progression-free survival (PFS) and overall survival (OS). The skeletal muscle index (SMI), subcutaneous fat index (SFI), and visceral fat index (VFI) at L3 were calculated using 3D Slicer.

Results

After a median follow-up of 29.7 months, 210 (78.3%) patients experienced disease progression and 185 (69.0%) died. The median overall survival (mOS) was 14.8 months, and the median progression-free survival (mPFS) was 7.2 months. The optimal cutoff value for WG was 4%, classifying patients into the WG ≥4% (n=87) and WG <4% (n=181) groups. WG ≥4% was associated with longer mOS (hazard ratio [HR], 0.225; P<0.001) and mPFS (HR, 0.449; P<0.001). WG ≥4% was also significantly associated with prolonged mOS in patients with stable disease (SD) (HR, 0.140; P<0.001) and progressive disease (PD) (HR, 0.045; P=0.003). Multivariate Cox regression analysis confirmed WG as an independent predictor of survival outcomes. SFI (r=0.34) and SMI (r=0.32) were moderately and positively correlated with OS.

Conclusions

WG ≥4% is associated with better survival outcomes in patients with stage IV GC undergoing combination immunotherapy, even in those with SD or PD.

INTRODUCTION

Antibodies targeting programmed death receptor 1 (PD-1) and programmed death ligand 1 (PD-L1) block signaling pathways between CD8⁺ T cells and both tumor cells and immunosuppressive cells, thereby restoring the ability of CD8⁺ T cells to recognize and kill tumor cells [1]. Based on several randomized controlled trials (RCTs), immunotherapy combined with chemotherapy and anti-human epidermal growth factor receptor 2 (HER2) therapy (for HER2-positive cases) has become the standard first-line treatment for advanced gastric cancer (GC), particularly in patients with PD-L1 combined positive score positivity [2,3,4,5].

Treatment efficacy is generally assessed using the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1) [6]. However, the distinct antitumor mechanisms of immunotherapy can produce atypical response patterns, such as pseudoprogression [7]. Across several tumor types, patients experiencing pseudoprogression have been shown to benefit from continued immunotherapy after imaging assessment and to have significantly longer survival than patients with true progressive disease (PD) [8,9,10]. Because pseudoprogression is difficult to distinguish from true PD on imaging, conventional RECIST 1.1 assessment may misclassify these patients and potentially lead to inappropriate treatment discontinuation. Therefore, new predictive markers are needed to more accurately assess true clinical benefit.

GC often causes mechanical obstruction in the digestive system and substantial tumor burden, which can lead to malnutrition and weight loss [11]. Therefore, nutritional status is an important indicator of disease severity and a critical prognostic factor for patient survival and quality of life [12]. Several clinical studies have shown that weight change is a prognostic factor for survival outcomes in patients with advanced GC [13,14,15,16,17]. A post hoc analysis of the phase III TAGS trial found that early weight loss of ≥3% was a significant adverse prognostic factor in patients with metastatic GC receiving third-line or later treatment [18]. However, the association between early weight change and survival outcomes in patients with stage IV GC receiving first-line combination immunotherapy remains unclear. Evaluating the relationship between weight gain (WG) and prognosis may provide additional clinical insight into treatment effectiveness and help determine more accurately whether immunotherapy should be continued.

MATERIALS AND METHODS

Patients

Patients were identified from the prospective GC immunotherapy clinical database established by the Department of General Surgery at Nanfang Hospital, Southern Medical University. A total of 268 patients with GC who received first-line PD-1 inhibitor-based combination therapy between October 2019 and October 2023 were included.

Patients were selected according to the following inclusion and exclusion criteria. Inclusion criteria were: 1) histologically confirmed gastric adenocarcinoma on pretreatment endoscopic biopsy; 2) stage IV GC confirmed by radiologic or laparoscopic evaluation (American Joint Committee on Cancer/Union for International Cancer Control 8th staging system); 3) receipt of a PD-1 inhibitor combined with chemotherapy and/or targeted therapy as first-line treatment; 4) presence of measurable lesions as defined by RECIST 1.1; and 5) good organ function. Exclusion criteria were: 1) a history of gastrectomy at any time before enrollment; 2) absence of abdominal computed tomography (CT) images within 30 days before initial treatment; 3) absence of early-treatment CT images within 30 days after the third or fourth treatment cycle; 4) a history of lumbar surgery affecting skeletal muscle or fat measurement; 5) other concurrent malignant tumors; and 6) active hematologic, inflammatory, or autoimmune diseases.

The study was approved by the Medical Ethics Committee of Nanfang Hospital, Southern Medical University (approval NFEC2024-458). The requirement for written informed consent was waived by the Ethics Committee because of the retrospective nature of the study and the use of de-identified clinical data. All procedures were conducted in accordance with the ethical standards of the research committees on human experimentation and the 1964 Declaration of Helsinki and its subsequent amendments [19].

Study assessments

Overall survival (OS) was defined as the time from initiation of PD-1 inhibitor-based combination therapy to death or the date of last follow-up. Progression-free survival (PFS) was defined as the time from initiation of PD-1 inhibitor-based combination therapy to objective tumor progression or death. Patients who had not experienced progression or death at the time of analysis were censored on the date of their last available tumor assessment. During treatment, all patients underwent CT evaluations every 2 to 3 months. Treatment responses were retrospectively assessed according to RECIST 1.1 [6]. Radiographic tumor assessments for disease progression continued even after PD-1 inhibitor therapy was discontinued because of toxicity or other reasons. The objective response rate (ORR) was defined as the proportion of patients who achieved a complete response (CR) or partial response (PR) as an early treatment outcome.

Assessment of nutritional support

Medically prescribed nutritional support during the early weight-change assessment period was retrospectively reviewed in the medical records. Routine oral food intake, including a normal daily diet, was not classified as specialized nutritional support. Nutritional support was categorized into three groups: no documented specialized nutritional support; oral nutritional supplementation or enteral nutrition (ONS/EN) only; and parenteral nutritional support (PN), with or without ONS/EN. ONS/EN included prescribed enteral nutritional formulas administered orally or through an enteral feeding tube. PN included lipid emulsions, amino acid preparations, and multicomponent glucose-amino acid-lipid preparations. Simple intravenous fluids or electrolyte/glucose infusions without amino acid or lipid components were not classified as specialized nutritional support. Parenteral nutritional preparations were not classified as total parenteral nutrition unless the medical records explicitly indicated that parenteral nutrition was the primary or exclusive nutritional source.

Assessment of ascites

Ascites during the early weight-change assessment period was retrospectively evaluated using CT reports and clinical records at the two body-weight measurement time points. Ascites was semi-quantitatively graded according to the Mayo Clinic ascites grading system as follows: grade 0, no ascites; grade 1, mild ascites; grade 2, moderate ascites; and grade 3, massive ascites [20]. Changes in ascites were classified as improved, stable, or worsened according to whether the ascites grade decreased, remained unchanged, or increased from baseline to the early assessment.

Image segmentation of body composition

CT images from 268 patients with stage IV GC were retrieved from the Picture Archiving and Communication System (PACS). Images were obtained within 30 days before the first treatment and within 30 days after the third or fourth treatment cycle. Cross-sectional CT images at the third lumbar level (L3) were selected and saved in DICOM format. In 3D Slicer version 5.6.2, the abdominal window was set to a window level of 50 Hounsfield units (HU) and a window width of 350 HU. Thresholding was then used to segment body composition in this region. The threshold for skeletal muscle segmentation was set between −29 and +150 HU and included the psoas, erector spinae, quadratus lumborum, internal and external oblique, and rectus abdominis muscles. The threshold range for subcutaneous fat was set between −190 and −30 HU, and that for visceral fat was set between −150 and −50 HU, with intraluminal fat in the abdominal cavity excluded. The software automatically calculated skeletal muscle area (SMA), subcutaneous fat area (SFA), and visceral fat area (VFA).

Evaluation of weight and body composition changes

Weight data were collected within 30 days before the first treatment (Weight^Pretreatment) and at the first early radiologic response assessment, which occurred within 30 days after the third or fourth treatment cycle (approximately 2 to 3 months after treatment initiation) (Weight^{Post-treatment}). WG was calculated using the following formula: $\frac{{\text{Weight}}^{\text{Post-treatment}}-{\text{Weight}}^{\text{Pretreatment}}}{{\text{Weight}}^{\text{Pretreatment}}}\text{×100\%}$. A WG greater than 0 indicated weight gain, whereas a WG less than 0 indicated weight loss. The threshold was determined to be 4.1% using the highest standardized log-rank statistic. To facilitate clinical use and select the cutoff associated with the lowest hazard ratio (HR), 4% was chosen as the optimal cutoff value, dividing patients into WG ≥4% and WG <4% groups for efficacy assessment.

SMA, SFA, and VFA were each divided by the square of the corresponding height to calculate skeletal muscle index (SMI), subcutaneous fat index (SFI), and visceral fat index (VFI) (cm²/m²), respectively. The prognostic nutritional index (PNI) was calculated using the formula: 5 × Lymphocyte Count (10⁹/L) + Serum Albumin (g/L) [21]. The change rate of each parameter was calculated using the following formula: $\frac{{\text{Parameters}}^{\text{Post-treatment}}-{\text{Parameters}}^{\text{Pretreatment}}}{{\text{Parameters}}^{\text{Pretreatment}}}$.

Statistical analyses

The Shapiro-Wilk test was used to assess the normality of continuous variables. Normally distributed variables were compared using the t-test; otherwise, the Mann-Whitney U test was used. Categorical variables were analyzed using the χ² test or Fisher’s exact test. The Kaplan-Meier method was used to generate survival curves, and between-group differences were compared using the log-rank test. Cox proportional hazards regression was used to calculate HRs and 95% confidence intervals (95% CIs). The proportional hazards assumption of the Cox model was tested using Schoenfeld residuals. A P>0.05 indicated that the residuals were independent of time and that the proportional hazards (PH) assumption was satisfied. Variables with a P<0.1 in univariate analysis were included in the multivariate analysis to identify potential risk factors for OS and PFS. G-computation based on the contsurvplot package was used to visualize the continuous association between WG and survival. Pearson correlation coefficients were used to quantify correlations between variables. Statistical analyses and graphical presentations were performed using R version 4.4.0 (R Foundation for Statistical Computing, Vienna, Austria). A 2-sided P-value of <0.05 was considered statistically significant.

RESULTS

Baseline characteristics

Between October 1, 2019, and October 31, 2023, 268 patients were included in this study. The median follow-up duration was 29.7 months. The median age was 55.6 years; 166 patients were male and 102 were female. The patients' clinicopathologic characteristics are presented in Table 1. Most patients (56.7%) had a normal weight. The median interval to imaging evaluation after early treatment was 2.9 months. The interval was not normally distributed in either WG group; the median was 2.9 months (interquartile range [IQR], 1.0) in the WG ≥4% group and 2.9 months (IQR, 0.9) in the WG <4% group, with no statistically significant difference between groups (Mann-Whitney U test, P=0.881).

Table 1. Patient baseline characteristics.

Variables		Overall (n=268)	WG		P-value
			≥4% (n=87)	<4% (n=181)
Age					0.17
	≤60 yr	156 (58.21)	45 (51.72)	111 (61.33)
	>60 yr	112 (41.79)	42 (48.28)	70 (38.67)
Age (yr)		55.6±12.0	56.5±12.8	55.1±11.6	0.12
Sex					0.33
	Male	166 (61.94)	58 (66.67)	108 (59.67)
	Female	102 (38.06)	29 (33.33)	73 (40.33)
BMI category					0.04
	Underweight (<18.5 kg/m2)	53 (19.78)	22 (25.29)	31 (17.13)
	Normal weight (18.5–23.9 kg/m2)	152 (56.72)	45 (51.72)	107 (59.12)
	Overweight (24–27.9 kg/m2)	58 (21.64)	16 (18.39)	42 (23.20)
	Obesity (≥28 kg/m2)	5 (1.87)	4 (4.60)	1 (0.55)
BMI (kg/m2)		21.47±3.22	21.32±3.46	21.55±3.11	0.07
ECOG PS					0.40
	0–1	260 (97.01)	86 (98.85)	174 (96.13)
	>1	8 (2.99)	1 (1.15)	7 (3.87)
CEA					0.04
	Normal	203 (75.75)	73 (83.91)	130 (71.82)
	Elevated	65 (24.25)	14 (16.09)	51 (28.18)
CA 19-9					0.29
	Normal	184 (68.66)	64 (73.56)	120 (66.30)
	Elevated	84 (31.34)	23 (26.44)	61 (33.70)
CA 72-4					0.32
	Normal	188 (70.15)	65 (74.71)	123 (67.96)
	Elevated	80 (29.85)	22 (25.29)	58 (32.04)
AFP					0.02
	Normal	207 (77.24)	67 (77.01)	140 (77.35)
	Elevated	34 (12.69)	6 (6.90)	28 (15.47)
	Unknown	27 (10.07)	14 (16.09)	13 (7.18)
Tumor location					0.02
	Lower	106 (39.55)	29 (33.33)	77 (42.54)
	Middle	52 (19.40)	17 (19.54)	35 (19.34)
	Upper	81 (30.22)	36 (41.38)	45 (24.86)
	Mixed	29 (10.82)	5 (5.75)	24 (13.26)
Lauren type					0.43
	Intestinal type	26 (9.70)	9 (10.34)	17 (9.39)
	Diffuse and mixed type	119 (44.40)	43 (49.43)	76 (41.99)
	Unknown	123 (45.90)	35 (40.23)	88 (48.62)
Borrmann criteria					0.59
	I	13 (4.85)	5 (5.75)	8 (4.42)
	II	9 (3.36)	4 (4.60)	5 (2.76)
	III	99 (36.94)	36 (41.38)	63 (34.81)
	IV	24 (8.96)	8 (9.20)	16 (8.84)
	Unknown	123 (45.90)	34 (39.08)	89 (49.17)
Histological grade					0.02
	Well or moderately differentiated	45 (16.79)	22 (25.29)	23 (12.71)
	Poorly differentiated	205 (76.49)	62 (71.26)	143 (79.01)
	Unknown	18 (6.72)	3 (3.45)	15 (8.29)
Depth of invasion					0.53
	≤cT3	26 (9.70)	11 (12.64)	15 (8.29)
	cT4a	116 (43.28)	36 (41.38)	80 (44.20)
	cT4b	126 (47.01)	40 (45.98)	86 (47.51)
Lymph node metastasis					0.32
	cN0	17 (6.34)	8 (9.20)	9 (4.97)
	cN1	45 (16.79)	17 (19.54)	28 (15.47)
	cN2	55 (20.52)	14 (16.09)	41 (22.65)
	cN3	151 (56.34)	48 (55.17)	103 (56.91)
Distant metastasis					0.93
	M0	61 (22.76)	19 (21.84)	42 (23.20)
	M1	207 (77.24)	68 (78.16)	139 (76.80)
PD-L1 status					0.17
	CPS <5	124 (46.27)	37 (42.53)	87 (48.07)
	CPS ≥5	139 (51.87)	50 (57.47)	89 (49.17)
	Unknown	5 (1.87)	0 (0.00)	5 (2.76)
HER2 status					0.81
	Negative	215 (80.22)	69 (79.31)	146 (80.66)
	Positive	45 (16.79)	16 (18.39)	29 (16.02)
	Unknown	8 (2.99)	2 (2.30)	6 (3.31)
MMR status					0.20
	dMMR	19 (7.09)	5 (5.75)	14 (7.73)
	pMMR	222 (82.84)	77 (88.51)	145 (80.11)
	Unknown	27 (10.07)	5 (5.75)	22 (12.15)
EBER status					0.79
	Negative	237 (88.43)	77 (88.51)	160 (88.40)
	Positive	19 (7.09)	7 (8.05)	12 (6.63)
	Unknown	12 (4.48)	3 (3.45)	9 (4.97)

Values are presented as mean ± standard deviation or number (%).

WG = weight gain; BMI = body mass index; ECOG PS = Eastern Cooperative Oncology Group performance status; CEA = carcinoembryonic antigen; CA = carbohydrate antigen; AFP = alpha-fetoprotein; PD-L1 = programmed death ligand 1; CPS = combined positive score; HER2 = human epidermal growth factor receptor 2; MMR = mismatch repair; dMMR = deficient mismatch repair; pMMR = proficient mismatch repair; EBER = Epstein-Barr virus-encoded small RNA.

We identified 181 patients (67.5%) with WG <4% and 87 patients (32.5%) with WG ≥4%. Except for body mass index (BMI) level, carcinoembryonic antigen (CEA), alpha-fetoprotein (AFP), tumor location, and histological grade, baseline demographic and disease characteristics were generally similar between the WG <4% and WG ≥4% groups. Compared with patients with WG <4%, those with WG ≥4% had a higher ORR (59.8% vs. 55.3%), a higher proportion of well or moderately differentiated tumors (25.3% vs. 12.7%), and a higher proportion with normal CEA levels (83.9%).

Treatment regimens and nutritional support

All patients received first-line chemotherapy combined with PD-1 inhibitor-based immunotherapy (Supplementary Table 1). In routine clinical practice, 10 mg of dexamethasone was administered intravenously with each intravenous chemotherapy infusion as standard antiemetic prophylaxis for chemotherapy-induced nausea and vomiting. In addition, 43 of the 45 patients with HER2-positive disease received trastuzumab-containing targeted therapy combined with chemotherapy and immunotherapy. The distributions of PD-1 inhibitor type, chemotherapy backbone, and trastuzumab use were comparable between the WG <4% and WG ≥4% groups, with no significant between-group differences in PD-1 inhibitor type (P=0.230), chemotherapy backbone (P=0.345), or trastuzumab use (P=0.468).

Specialized nutritional support during the early weight-change assessment period was also reviewed (Supplementary Table 1). Overall, 119 patients (44.4%) had no documented specialized nutritional support, 132 (49.3%) received ONS/EN only, and 17 (6.3%) received PN support with or without ONS/EN. No patient received continuous total parenteral nutrition as the exclusive nutritional source throughout the early weight-change assessment period. The distribution of nutritional support did not differ significantly between the WG <4% and WG ≥4% groups (P=0.153). ONS/EN was the predominant nutritional intervention in both groups, whereas PN support was uncommon.

Determining the optimal cutoff value for WG

The optimal WG cutoff was determined to be 4.1% using the highest standardized log-rank statistic (Supplementary Fig. 1). However, for ease of clinical application and because the most significant OS difference occurred at a WG of 4% (Supplementary Table 2), we set the WG cutoff at 4%. Consequently, patients were divided into 2 groups: WG ≥4% and WG <4%. The distribution of WG according to survival status indicated that higher WG was associated with a lower risk of progression or death (Supplementary Fig. 2).

Correlation between WG and prognosis

Among the 268 patients included in this study, the median progression-free survival (mPFS) was 7.2 months, with 210 patients (78.4%) experiencing disease progression (Fig. 1A), and the median overall survival (mOS) was 14.8 months, with 185 patients (69.0%) dying (Fig. 1B). Compared with the WG <4% group, the WG ≥4% group had lower mortality (47.1% vs. 79.6%) and disease progression rates (65.5% vs. 84.5%) and higher OS and PFS rates (Supplementary Table 3). The mOS was 30.1 months in the WG ≥4% group, significantly longer than the 11.3 months in the WG <4% group (HR, 0.225; P<0.001) (Fig. 1B). Similarly, the mPFS was 14.5 months in the WG ≥4% group, longer than the 5.6 months in the WG <4% group (HR, 0.449; P<0.001) (Fig. 1A).

Fig. 1. Kaplan-Meier survival curves according to WG. (A) PFS in the WG <4% and WG ≥4% groups. (B) OS in the WG <4% and WG ≥4% groups.

WG = weight gain; PFS = progression-free survival; OS = overall survival; CI = confidence interval; HR = hazard ratio; NA = not available.

Univariate Cox regression analysis showed that WG, tumor location, depth of invasion, lymph node metastasis, distant metastasis, CEA levels, and carbohydrate antigen (CA) 724 levels were significantly correlated with OS (Table 2). Multivariate Cox regression analysis indicated that WG, modeled as a continuous variable, was an independent predictor of OS (HR, 0.897; P<0.001) and PFS (HR, 0.941; P<0.001). Distant metastasis (HR, 1.701; P=0.007) and depth of invasion (HR, 1.781; P=0.040) retained independent prognostic value for PFS, whereas lymph node metastasis (cN stage) did not. None of the T, N, or M factors remained independently associated with OS in the multivariate model. The proportional hazards assumption for the Cox regression model was satisfied (Supplementary Fig. 3).

Table 2. Univariate and multivariate Cox regression analyses of WG and clinicopathologic characteristics for OS and PFS.

Variable		OS				PFS
		Univariate analysis		Multivariate analysis		Univariate analysis		Multivariate analysis
		HR (95% CI)	P-value	HR (95% CI)	P-value	HR (95% CI)	P-value	HR (95% CI)	P-value
Age (yr) (>60 vs. ≤60)		0.904 (0.673–1.215)	0.504			0.915 (0.694–1.206)	0.528
Sex (female vs. male)		0.970 (0.721–1.306)	0.843			0.914 (0.690–1.209)	0.528
BMI category
	Normal
	Underweight	0.883 (0.601–1.296)	0.524			1.174 (0.823–1.676)	0.377
	Overweight	1.177 (0.819–1.690)	0.378			1.092 (0.776–1.536)	0.614
	Obesity	0.988 (0.402–2.429)	0.979			1.278 (0.520–3.140)	0.592
ECOG PS (>1 vs. 0–1)		1.606 (0.711–3.628)	0.255			1.144 (0.505–2.593)	0.747
WG		0.898 (0.879–0.918)	<0.001	0.897 (0.877–0.918)	<0.001	0.936 (0.919–0.954)	<0.001	0.941 (0.923–0.958)	<0.001
Tumor location
	Upper
	Middle	1.056 (0.693–1.607)	0.801	0.665 (0.424–1.042)	0.075	1.546 (1.043–2.290)	0.030	1.117 (0.743–1.680)	0.595
	Lower	0.995 (0.697–1.420)	0.977	0.721 (0.492–1.058)	0.095	1.117 (0.798–1.564)	0.519	0.866 (0.613–1.225)	0.417
	Mixed	1.880 (1.156–3.059)	0.011	1.451 (0.869–2.423)	0.155	1.478 (0.928–2.352)	0.100	1.118 (0.697–1.793)	0.645
Depth of invasion
	≤cT3
	cT4a	2.203 (1.227–3.953)	0.008	1.771 (0.956–3.280)	0.069	1.835 (1.063–3.168)	0.029	1.781 (1.025–3.094)	0.040
	cT4b	1.654 (0.919–2.977)	0.093	1.231 (0.640–2.369)	0.534	1.478 (0.856–2.553)	0.161	1.521 (0.868–2.665)	0.143
Lymph node metastasis
	cN0
	cN1	1.529 (0.726–3.222)	0.264	1.279 (0.591–2.768)	0.532	0.973 (0.511–1.852)	0.933
	cN2	1.582 (0.758–3.302)	0.222	0.819 (0.381–1.764)	0.611	0.972 (0.518–1.823)	0.930
	cN3	1.904 (0.964–3.758)	0.064	1.307 (0.641–2.664)	0.461	1.195 (0.674–2.119)	0.542
Distant metastasis (M1 vs. M0)		1.434 (0.988–2.081)	0.058	1.354 (0.887–2.068)	0.161	1.793 (1.250–2.570)	0.001	1.701 (1.156–2.504)	0.007
CEA level (elevated vs. normal)		1.573 (1.138–2.175)	0.006	1.257 (0.882–1.791)	0.206	1.250 (0.916–1.705)	0.160
CA 19-9 level (elevated vs. normal)		1.235 (0.910–1.676)	0.176			1.448 (1.090–1.922)	0.011	1.192 (0.878–1.618)	0.261
CA 72-4 level (elevated vs. normal)		1.473 (1.085–1.999)	0.013	1.261 (0.908–1.750)	0.166	1.515 (1.134–2.022)	0.005	1.276 (0.945–1.724)	0.112
Nutritional support
	No documented specialized nutritional support
	ONS/EN only	0.957 (0.710–1.291)	0.776			1.043 (0.789–1.379)	0.767
	PN support with or without ONS/EN	0.825 (0.439–1.551)	0.551			0.743 (0.407–1.357)	0.334
HER2 status
	Negative
	Positive	0.756 (0.502–1.138)	0.180			0.851 (0.583–1.242)	0.403
MMR status
	pMMR
	dMMR	1.158 (0.657–2.040)	0.613			1.234 (0.699–2.176)	0.469
PD-L1 status
	CPS <5
	CPS ≥5	0.838 (0.625–1.124)	0.239			0.997 (0.758–1.313)	0.985
EBER status
	Negative
	Positive	1.039 (0.590–1.829)	0.894			0.922 (0.525–1.619)	0.778

Associations of all variables with prognosis were analyzed using two-sided Cox proportional hazards regression.

WG = weight gain; OS = overall survival; PFS = progression-free survival; HR = hazard ratio; CI = confidence interval; BMI = body mass index; ECOG PS = Eastern Cooperative Oncology Group performance status; CEA = carcinoembryonic antigen; CA = carbohydrate antigen; ONS = oral nutritional supplementation; EN = enteral nutrition; PN = parenteral nutrition; HER2 = human epidermal growth factor receptor 2; MMR = mismatch repair; pMMR = proficient mismatch repair; dMMR = deficient mismatch repair; PD-L1 = programmed death ligand 1; CPS = combined positive score; EBER = Epstein-Barr virus-encoded small RNA.

Across models adjusted for potential confounders, WG was consistently associated with a reduced risk of death and disease progression. Similar results were observed when WG was modeled by quartiles (Supplementary Table 4). After adjustment for potential confounders (Model 2), the HR for death decreased sequentially with increasing WG, from 1 to 0.527, 0.386, and 0.141. The P-value for trend was <0.001, indicating a statistically significant downward trend. These findings suggest that among patients with GC treated with PD-1 inhibitor-based combination therapy, the risk of death decreases as WG increases. G-computation was performed to assess OS and PFS probabilities at different time points. The analysis showed a significant positive correlation between WG and the probabilities of OS and PFS (Fig. 2A and B). The continuous survival probability surface plot provides an integrated visualization of how WG modulates survival over time (Fig. 2C and D). At any given time point, higher WG (red regions) was consistently associated with greater survival probabilities than lower WG (blue regions). Similarly, following a given WG level over time showed that although survival probabilities declined in all patients, the decline was markedly slower at higher WG levels, as indicated by the persistent red shading over time.

Fig. 2. Survival probability according to WG. (A) OS probability at different time points as a function of WG. (B) PFS probability at different time points as a function of WG. (C) Continuous-color survival region plot showing the relationship between OS probability and WG over time. (D) Continuous-color survival region plot showing the relationship between PFS probability and WG over time.

WG = weight gain; OS = overall survival; PFS = progression-free survival.

WG as a prognostic factor across different treatment responses

For patients receiving PD-1 inhibitor-based combination therapy, early efficacy assessment according to RECIST 1.1 (within 30 days after the third or fourth treatment cycle) is a robust prognostic indicator. The mOS was 19.5 months for patients with CR/PR, 11.3 months for those with stable disease (SD), and 7.2 months for those with PD (Fig. 3A). The corresponding mPFS values were 8.9, 7.1, and 2.6 months, respectively (Fig. 3B).

Fig. 3. Kaplan-Meier survival curves according to RECIST 1.1. (A) OS and (B) PFS according to treatment response.

RECIST 1.1 = Response Evaluation Criteria in Solid Tumors version 1.1; OS = overall survival; PFS = progression-free survival; CR = complete response; PR = partial response; SD = stable disease; PD = progressive disease; CI = confidence interval; HR = hazard ratio.

WG was associated not only with overall prognosis but also with prognosis across different early treatment responses. Among patients with CR/PR, the mOS was 36.2 months in the WG ≥4% group compared with 14.7 months in the WG <4% group (HR, 0.199; P<0.001) (Fig. 4A). Among patients with SD, the mOS was 24.4 months in the WG ≥4% group compared with 9.0 months in the WG <4% group (HR, 0.140; P<0.001) (Fig. 4C). Among patients with PD, the mOS was 12.7 months in the WG ≥4% group compared with 5.2 months in the WG <4% group (HR, 0.045; P<0.001) (Fig. 4E). In the CR/PR, SD, and PD groups, WG ≥4% reduced the risk of death by 80%, 86%, and 96%, respectively, indicating that weight gain significantly lowers the risk of death, particularly in patients with less favorable early imaging assessments. Similarly, PFS was significantly longer in the WG ≥4% group than in the WG <4% group among patients with different early treatment responses, except those with PD (P=0.113) (Fig. 4B, D, and F).

Fig. 4. Kaplan-Meier survival curves according to WG in patients with different treatment responses. (A) OS and (B) PFS in patients with CR/PR stratified by WG; (C) OS and (D) PFS in patients with SD stratified by WG; and (E) OS and (F) PFS in patients with PD stratified by WG.

WG = weight gain; OS = overall survival; PFS = progression-free survival; CR = complete response; PR = partial response; SD = stable disease; PD = progressive disease; CI = confidence interval; HR = hazard ratio; NA = not available.

Additionally, when analyses were stratified by potential risk factors, WG appeared to be a protective factor in most subgroups, supporting its independent association with prognosis (Supplementary Fig. 4). Significant interactions with WG were observed only for CA 724 (P for interaction=0.025) and BMI status (P for interaction=0.017).

Changes in body composition correlate with WG and positively impact OS

To further clarify the nature of early weight change, we assessed ascites status at the two body-weight measurement time points and performed CT-based body composition analysis using L3-level images obtained before treatment and at the early post-treatment assessment (Supplementary Table 5, Fig. 5). At baseline, 162 patients (60.4%) had no ascites, 102 (38.1%) had mild ascites, 2 (0.7%) had moderate ascites, and 2 (0.7%) had massive ascites. At the early assessment, 172 patients (64.2%) had no ascites, 85 (31.7%) had mild ascites, 8 (3.0%) had moderate ascites, and 3 (1.1%) had massive ascites. The distribution of ascites status did not differ significantly between the WG ≥4% and WG <4% groups at baseline (P=0.388) or at the early assessment (P=0.237). We further evaluated paired changes in ascites between the 2-time points. Ascites improved in 51 patients (19.0%), remained stable in 169 (63.1%), and worsened in 48 (17.9%). Although the WG ≥4% group had numerically higher proportions of ascites improvement and stability than the WG <4% group (21.8% vs. 17.7% and 67.8% vs. 60.8%, respectively) and a numerically lower proportion of ascites worsening (10.3% vs. 21.5%), the distribution of ascites change did not differ significantly between groups (P=0.078) (Supplementary Table 6).

Fig. 5. Measurement of skeletal muscle (red), subcutaneous fat (green), and visceral fat (yellow) on CT images. (A) Pretreatment and (B) post-treatment CT images of patient 1. (C) Pretreatment and (D) post-treatment CT images of patient 2.

CT = computed tomography.

To better characterize changes in body composition, SFA, VFA, and SMA were segmented and measured on L3 CT images before and after early treatment. WG was significantly correlated with SFI, VFI, and SMI percentage change (Fig. 6), with particularly strong correlations with the SFI (r=0.66, P<0.001) and VFI (r=0.55, P<0.001) percentage change. In addition, the SFI (r=0.34, P<0.001) and SMI (r=0.32, P<0.001) percentage change were significantly associated with OS, whereas the VFI change rate (r=0.17, P=0.008) and PNI change rate (r=−0.08, P=0.21) showed weaker or nonsignificant correlations with OS (Supplementary Fig. 5). Overall, WG was primarily associated with increases in subcutaneous fat, visceral fat, and skeletal muscle, suggesting that these body composition changes may positively impact patient OS.

Fig. 6. Correlation heatmap of WG and body composition percentage change.

WG = weight gain; PNI = prognostic nutritional index; VFI = visceral fat index; SFI = subcutaneous fat index; SMI = skeletal muscle index.

The values indicate correlation coefficients. Asterisks indicate statistical significance of the correlations: ^**P<0.01 and ^***P<0.001.

DISCUSSION

RECIST 1.1 is the recognized standard for assessing treatment response in solid tumors. However, pseudoprogression may occur during combination immunotherapy, potentially limiting the ability of RECIST 1.1 to accurately reflect true efficacy and increasing the risk of misclassification. In this retrospective analysis of patients with stage IV GC treated with PD-1 inhibitor-based combination therapy, WG ≥4% was strongly associated with longer mOS and mPFS.

Weight loss is a hallmark of cancer cachexia and is closely linked to patients’ quality of life and survival [22]. It is also an independent prognostic factor in patients with advanced GC [17], lung cancer [23], and colorectal cancer [24]. In patients with GC, weight changes generally occur earlier and are more widespread than other prognostic indicators; weight is also easier and less costly to monitor, offering practical clinical advantages. A retrospective analysis of a phase III clinical study found that weight loss ≥3% during the first cycle of chemotherapy was a negative prognostic factor for OS in patients with advanced GC and gastroesophageal junction adenocarcinoma (9.7 vs. 11.7 months) [13]. Although weight change has been recognized as a key prognostic indicator in patients with GC after chemotherapy or targeted therapy [13,16,17,18,25] and is closely linked to immune-inflammatory states in patients with cancer [26], its association with prognosis after combination immunotherapy remains uncertain.

A retrospective study showed that the weight change rate during the first month of palliative chemotherapy in patients with advanced GC was more strongly correlated with OS than the weight change rate at the end of chemotherapy, the average monthly weight change rate, or the initial BMI [16]. The present study standardized the rate of weight change before and after treatment, improving comparability among patients with different baseline weights and capturing dynamic weight changes during treatment. The optimal cutoff value for WG was identified as 4% using the highest standardized log-rank statistic, maximizing discrimination between patient prognoses. Analysis of WG as a continuous variable revealed a significant dose-response relationship, with each 1% gain conferring an approximately 10% reduction in mortality risk (HR, 0.897). However, exceeding the 4% threshold was associated with a substantially greater survival benefit (OS: HR, 0.225), suggesting a nonlinear, threshold-based protective effect. WG ≥4% significantly prolonged mPFS and mOS. As follow-up increased, the survival advantage among patients with WG ≥4% also increased: the difference in 1-year OS rates was 40% (86% vs. 46%), increasing to 52% at 2 years (65% vs. 13%) and becoming even more pronounced at 3 years (42% vs. not applicable), indicating a significant impact of weight gain on long-term prognosis. In multivariate analysis, early WG remained the primary independent prognostic factor for both OS and PFS, whereas baseline BMI did not show comparable prognostic utility. Notably, although the prognostic utility of most conventional staging factors was attenuated, cM1 and cT4a retained independent associations with PFS. These findings highlight that WG captures a distinct prognostic dimension integrating effective tumor control, cachexia reversal, and treatment tolerance. In addition, WG showed strong discriminatory power across subgroups with different clinical characteristics.

Tumor diameter is a direct indicator of treatment efficacy and is strongly associated with patient prognosis. RECIST 1.1 has become the standard for assessing solid tumor responses. When a patient meets the criteria for PD, treatment is considered unsuccessful and a change in therapy is required. Four-year follow-up results from the CheckMate-649 study highlighted the prognostic importance of initial treatment response: at week 18, the mOS for patients achieving CR/PR, SD, and PD was 23.3, 13.0, and 8.6 months, respectively [27]. The present study showed a similar trend: patients with initial treatment evaluations of CR/PR, SD, and PD had mOS values of 19.5, 11.3, and 7.2 months, respectively.

However, RECIST 1.1 has limitations in the setting of immunotherapy. Because immunotherapy activates antitumor immune mechanisms [28], it can lead to atypical tumor responses such as hyperprogression [29], dissociated responses [30], and pseudoprogression [31], which challenge conventional standards for evaluating cancer treatment efficacy. Pseudoprogression is generally defined as a temporary increase in tumor size or the emergence of new lesions during treatment despite a stable or improving clinical condition [32]. Two main mechanisms have been proposed: first, that the immune response has not yet been activated [33]; and second, that immune-system reactivation leads to immune-cell infiltration, temporarily increasing inflammation and tumor burden [31]. Pseudoprogression was first reported in patients with melanoma receiving cytotoxic T-lymphocyte antigen 4 inhibitors [34] and has since been observed across multiple tumor types, with an incidence of 2.8% to 15.8% [35]. In a retrospective study of 542 patients with advanced non-small cell lung cancer treated with nivolumab [36], OS was significantly longer in the pseudoprogression group than in the true-PD group (mOS: not reached vs. 6.4 months). Pseudoprogression has also been reported in patients with gastrointestinal cancers [37]. If pseudoprogression is misdiagnosed as PD, patients may miss potential benefits from continued immunotherapy. In recent years, new immunotherapy response criteria and imaging tracers have been introduced to distinguish true PD from pseudoprogression, including immune-related RECIST [31], immune-modified RECIST [38], immune-cell receptor tracers (such as ⁸⁹Zr-Lumretuzumab [39]), and amino acid tracers (such as ¹¹C-methyl-L-methionine [40]). However, the broader clinical use of these approaches remains limited.

The impact of weight gain on survival across different treatment responses after combination immunotherapy has not been fully characterized. This study used weight change, which is easy to monitor clinically, as an evaluation measure and found that WG, in addition to RECIST 1.1 assessment, could further distinguish survival outcomes and serve as a favorable prognostic indicator in patients with advanced GC. The improvement in OS with WG ≥4% was most pronounced in patients with CR/PR, with an mOS difference of 21.5 months (36.2 vs. 14.7 months). This finding suggests that early treatment response plays a key role in prognosis and that WG may enhance the treatment effect. In patients with SD, WG ≥4% also provided favorable prognostic information (24.4 vs. 9.0 months). Furthermore, among patients with PD, those with WG ≥4% had significantly longer mOS than those with WG <4% (12.7 vs. 5.2 months). Thus, routine weight monitoring may help identify patients with unfavorable early imaging findings but better-than-expected prognosis, complementing RECIST 1.1 and providing a more comprehensive basis for prognostic assessment.

Although the pathophysiologic mechanisms underlying the relationship between WG and improved prognosis remain unclear, they can be preliminarily explored using several nutrition- and body composition-related indicators. In patients with advanced GC, body-weight change may be influenced not only by nutritional and metabolic status but also by tumor burden, ascites, tissue edema, and treatment-related fluid administration. Malignant ascites is a common manifestation of advanced gastrointestinal malignancies and may cause weight gain [41], potentially confounding the interpretation of body-weight changes. To address this issue, we retrospectively assessed ascites status at the two body-weight measurement time points. Ascites worsening was not more frequent in the WG ≥4% group than in the WG <4% group, and the distribution of ascites change did not differ significantly between groups. These findings suggest that early WG was unlikely to be primarily attributable to worsening ascites during the early weight-change assessment period.

Skeletal muscle mass and adipose tissue are important components of cancer-related weight changes [22,42], and the cross-sectional areas of fat and muscle at L3 can effectively predict total body fat and muscle mass [43,44]. Lee et al. [45] found that among patients with advanced GC receiving palliative chemotherapy, those with sarcopenia had significantly shorter mOS than those without sarcopenia (6.8 vs. 10.3 months), and sarcopenia and low serum albumin were independent predictors of poor prognosis. However, among patients with advanced GC receiving dual blockade with PD-1 and HER2, high SFA was independently associated with longer OS, whereas sarcopenia was not an independent prognostic factor for OS [46]. In the present study, increases in SMI, SFI, and VFI were significantly correlated with both WG and improved OS. This finding suggests that early WG captures a treatment-induced shift toward improved anabolic status and a better systemic condition, a dynamic process that static baseline BMI cannot represent.

Overall, early weight gain can serve as an important predictor of prognosis in patients with advanced GC receiving combination immunotherapy. When assessed alongside RECIST 1.1, WG ≥4% can help identify patients with a better prognosis, particularly those with unfavorable early imaging findings. Therefore, for patients with SD or PD early in treatment who also experience substantial weight gain, maintaining the current treatment and appropriately extending the observation period before reassessing efficacy may be reasonable. This finding provides a useful complement to conventional RECIST 1.1 assessment and offers an additional perspective for more accurately evaluating immunotherapy efficacy.

First, the optimal cutoff value for weight gain was identified using a univariable approach (the highest standardized log-rank statistic). Although this is a common and practical method for exploratory cut-point determination in clinical research, it carries a risk of overfitting and potentially overestimating the effect size. Nevertheless, the robustness of the 4% cutoff was supported by its persistent significance in multivariable analyses. External validation in an independent cohort remains essential to confirm the generalizability of these findings. Second, detailed information on glucocorticoid use for the management of immune-related adverse events was not systematically collected. Because glucocorticoids may influence appetite, fluid retention, fat distribution, and body weight, residual confounding from corticosteroid exposure cannot be completely excluded. Finally, although ascites was retrospectively assessed using CT reports at the 2 body-weight measurement time points, data on peripheral edema were not systematically available, and ascites was assessed semi-quantitatively rather than volumetrically. Therefore, residual confounding from incompletely measured corticosteroid exposure and fluid retention cannot be completely excluded.

SUPPLEMENTARY MATERIALS

Supplementary Table 1

Treatment regimens and nutritional support

Supplementary Table 2

HRs for different WG cutoff values in predicting OS

Supplementary Table 3

Comparison of OS and PFS rates based on WG

Supplementary Table 4

HRs (95% CIs) for OS and PFS based on continuous WG and WG categorized by quartile

Supplementary Table 5

Ascites status at baseline and early assessment

Supplementary Table 6

Changes in ascites status according to WG group during the early weight-change assessment period

Supplementary Fig. 1

Selection of the optimal cutoff value for WG. (A) Histogram showing the density distributions of the WG <4% and WG ≥4% groups based on the optimal cutoff value. (B) Scatter plot showing the standardized log-rank statistic for each WG cutoff value. The cutoff value was determined using two-sided maximally selected rank statistics.

Supplementary Fig. 2

Distribution of WG according to survival status. (A) Distribution of WG. (B) Relationship between WG distribution and OS status. (C) Relationship between WG distribution and PFS status.

Supplementary Fig. 3

PH assumption test for the multivariate Cox regression model for OS.

Supplementary Fig. 4

Subgroup analysis of the association between WG group and OS.

Supplementary Fig. 5

Correlations between body composition measures and OS. (A) Correlation between PNI change rate and OS. (B) Correlation between SFI change rate and OS. (C) Correlation between VFI change rate and OS. (D) Correlation between SMI change rate and OS.