Comparative study of 18F-FAPI-04 and 18F-FDG PET/CT in the initial staging and postoperative metastasis detection of gastric cancer

Abstract

BACKGROUND

Accurate clinical staging is fundamental to the precision management of gastric cancer (GC). Nonetheless, the clinical performance of ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) positron emission tomography (PET)/computed tomography (CT) in GC is limited by relatively low sensitivity, particularly for lymph node metastasis (LNM) and peritoneal metastasis (PM).

AIM

To compare ¹⁸F-fibroblast activation protein inhibitor (FAPI)-04 PET/CT with ¹⁸F-FDG PET/CT in patients with newly diagnosed GC and in those with suspected recurrence or metastasis after curative surgery.

METHODS

Twenty-one patients were prospectively enrolled, including 11 with newly diagnosed GC and 10 with suspected postoperative recurrence or metastasis. Each patient underwent both ¹⁸F-FAPI-04 and ¹⁸F-FDG PET/CT within 7 days. Lesions were verified by pathology and/or imaging follow-up. Diagnostic performance and quantitative parameters were compared, including maximum standardized uptake value (SUVmax), tumor-to-background ratio (TBR), FDG-derived metabolic tumor volume (MTV) and total lesion glycolysis (TLG), and FAPI-derived FAPI-avid tumor volume (FTV), and total lesion fibroblast activation protein expression (TLF). The peritoneal cancer index (PCI) was assessed for diffuse peritoneal disease.

RESULTS

Among the 21 patients (237 lesions in total), 11 (52.4%, 11/21) had newly diagnosed GC, and 10 (47.6%, 10/21) had suspected local recurrence or metastasis after surgery; only 1 patient (4.8%, 1/21) had suspected local recurrence. For primary tumors, ¹⁸F-FAPI-04 demonstrated higher uptake and contrast than ¹⁸F-FDG (SUVmax: 11.6 vs 6.5, P = 0.036; TBR: 11.5 vs 5.1, P = 0.009) and yielded higher tracer-specific total-lesion burden metrics (TLF vs TLG: 260.5 vs 57.9, P = 0.036). For LNMs, quantitative uptake and derived parameters were comparable between tracers (SUVmax: 6.6 vs 6.5, P = 0.703; TBR: 5.4 vs 6.4, P = 0.65; FTV vs MTV: 3.1 vs 1.9, P = 0.321; TLF vs TLG: 8.3 vs 6.3, P = 0.654). For PM, ¹⁸F-FAPI-04 showed substantially higher uptake and volumetric burden in nodular disease (SUVmax: 7.1 vs 3.2, P = 0.002; TBR: 7.4 vs 2.4, P = 0.002; FTV vs MTV: 13.2 vs 1.2, P = 0.003; TLF vs TLG: 39.3 vs 3.3, P = 0.004). In diffuse PM, ¹⁸F-FAPI-04 yielded a higher PCI score and higher SUVmax than ¹⁸F-FDG (PCI: 14 vs 2.5, P = 0.012; SUVmax: 8.0 vs 4.7, P = 0.001). For bone metastases, ¹⁸F-FAPI-04 also demonstrated higher uptake and contrast (SUVmax: 11.1 vs 5.5, P < 0.001; TBR: 10.4 vs 5.6, P < 0.001) and differed in tracer-specific total-lesion metrics (TLF vs TLG: 7.7 vs 9.5, P = 0.045). Among the 11 patients with primary GC, the sensitivity of ¹⁸F-FAPI-04 PET/CT for primary tumor detection was 100% compared with 90.9% for ¹⁸F-FDG PET/CT. For detection of LNM and PM, sensitivities were 100% and 100% with ¹⁸F-FAPI-04, vs 75% and 50% with ¹⁸F-FDG, respectively.

CONCLUSION

In this pilot study, we observed trends of higher tracer uptake and lesion-to-background contrast with ¹⁸F-FAPI-04 PET/CT compared to ¹⁸F-FDG PET/CT in primary GC, PMs, and bone metastases. Additionally, ¹⁸F-FAPI-04 PET/CT showed trends of a greater number of metastatic lesions overall, with higher detection yields particularly for PM, suggesting its potential value for initial staging and postoperative assessment in GC.

Key Words: ¹⁸F-fibroblast activation protein inhibitor-04; ¹⁸F-fluorodeoxyglucose; Positron emission tomography/computed tomography; Gastric cancer; Cancer-associated fibroblasts

Core Tip: This prospective pilot study compares ¹⁸F-fibroblast activation protein inhibitor-04 and ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography in 21 gastric cancer (GC) patients. ¹⁸F-fibroblast activation protein inhibitor-04 shows higher uptake, contrast and tumor-burden estimates in primary GC, peritoneal and bone metastases, and higher sensitivity in detecting lymph node and peritoneal metastases, suggesting its potential value for initial staging and postoperative assessment in GC.

INTRODUCTION

Gastric cancer (GC) is among the most common malignancies worldwide and remains a leading cause of cancer-related mortality. As early clinical manifestations are often nonspecific, many patients are diagnosed at an advanced stage; globally, GC is the fifth leading cause of cancer death[1]. Cancer care has shifted toward a multimodal precision medicine system, in which targeted agents and immunotherapy, rather than conventional cytotoxic regimens, increasingly guide treatment selection[2]. In parallel, imaging is no longer confined to anatomical localization and treatment-response monitoring. For heterogeneous tumors like GC, management increasingly depends on resolving the underlying tumor biology. Thus, functional and molecular imaging and radiomics, integrated with genetic, pathological, and treatment-response data, are essential for individualized care in GC, spanning early detection, accurate staging, noninvasive evaluation of therapeutic targets, and drug resistance monitoring. ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) positron emission tomography (PET)/computed tomography (CT) is widely used for oncologic staging, treatment response assessment, and surveillance for recurrence across a wide range of malignancies. In GC, however, FDG uptake is influenced by histopathological subtype and physiological gastric activity, thereby limiting diagnostic performance, particularly for detecting peritoneal metastasis (PM)[3]. These limitations underscore the need for more effective molecular tracers that can enhance PET/CT-based assessment of GC and metastatic spread, supporting precision management.

Fibroblast activation protein (FAP), a characteristic marker of cancer-associated fibroblasts (CAFs), is highly expressed in the tumor stroma of many epithelial cancers but shows low or absent expression in most normal tissues[4]. On this basis, PET/CT and PET/magnetic resonance imaging using FAP inhibitors (FAPI) labeled with ⁶⁸Ga have demonstrated promising clinical value in GC[5,6]. Recent studies suggest that ¹⁸F-FAPI-74 PET/CT may offer advantages for the diagnosis and staging of GC[7,8]; however, evidence regarding ¹⁸F-FAPI-04 PET/CT in GC remains limited and warrants further validation. Therefore, this single-center prospective study aimed to compare ¹⁸F-FAPI-04 PET/CT with ¹⁸F-FDG PET/CT for the initial evaluation of newly diagnosed GC and for postoperative assessment of suspected metastatic lesions after gastrectomy.

MATERIALS AND METHODS

Patients

We enrolled patients who underwent paired ¹⁸F-FAPI-04 PET/CT and ¹⁸F-FDG PET/CT at the Affiliated Hospital of Jiangnan University between February 2023 and March 2024, including patients with newly diagnosed GC for initial staging and postoperative patients with suspected recurrence or metastatic disease. Inclusion criteria were as follows: (1) Treatment-naive patients with pathologically confirmed GC or clinically suspected GC; and (2) Patients who had undergone curative gastrectomy with follow-up findings suggestive of recurrence/metastasis on imaging and/or elevated tumor markers. Exclusion criteria were shown below: (1) Failure to complete paired ¹⁸F-FAPI-04 PET/CT and ¹⁸F-FDG PET/CT within 7 days (i.e., ¹⁸F-FDG PET/CT not performed and/or an inter-scan interval ≥ 7 days); (2) Incomplete clinical or follow-up data; (3) Presence of another primary malignancy at the time of evaluation; or (4) Presence of severe hepatic or renal insufficiency. All participants provided written informed consent. This prospective study was approved by the institutional ethics committee, approval No. LSX2022043 and registered at ClinicalTrials.gov (NCT05691894).

Radiotracer synthesis

Both radiotracers were prepared in-house in the Department of Nuclear Medicine, Affiliated Hospital of Jiangnan University. ¹⁸F was produced using an HM-10 cyclotron (Sumitomo Heavy Industries,Ltd., Japan). The NOTA-FAPI-04 precursor was provided by the Jiangsu Institute of Nuclear Medicine. ¹⁸F-AlF-NOTA-FAPI-04 was synthesized using an automated FASTLab^TM2 module (GE Healthcare, United States), and ¹⁸F-FDG was synthesized using a PET-¹⁸F-FDG-IT-N “four-in-one” synthesis module. Quality control demonstrated radiochemical purity > 95% for both ¹⁸F-FAPI-04 and ¹⁸F-FDG; both products met sterility requirements and standards for human injection.

PET/CT acquisition protocol

All ¹⁸F-FAPI-04 and ¹⁸F-FDG PET/CT examinations were performed on a Siemens Biograph 64 TruePoint PET/CT system (Siemens Healthineers, Germany), and the two scans for each patient were completed within 7 days. No special preparation was required for ¹⁸F-FAPI-04 PET/CT. For ¹⁸F-FDG PET/CT, patients fasted for at least 4 hours, and blood glucose was required to be < 11.1 mmol/L. The administered activities of both tracers were weight-based 3.7 MBq (0.1mCi)/kg. For ¹⁸F-FDG PET/CT, images were acquired with an uptake time of 65.7 ± 9.0 minutes (range: 50-78 minutes) following an injection of 294.15 ± 64.75 MBq (range: 165.0-416.62 MBq). For ¹⁸F-FAPI-04 PET/CT, after an injection of 287.12 ± 44.03 MBq (range: 197.95-366.3 MBq), imaging was performed with an uptake time of 68.9 ± 5.3 minutes (range: 60-75 minutes). For both tracers, CT was acquired first, followed by PET. CT parameters were: Tube voltage 120 kV, tube current 170 mA, pitch 0.8, and slice thickness 5 mm. PET acquisition time was 1.5 minutes per bed position for the body and 3.0 minutes per bed position for the head. PET data were corrected using CT-based attenuation correction, and images were reconstructed using an iterative method. All datasets were transferred to a Syngo 6.5 (2012A) workstation for post-processing.

¹⁸F-FDG and ¹⁸F-FAPI-04 PET/CT image analysis

All images were independently interpreted by two nuclear medicine physicians (Xue XQ, 13 years of experience; Li QB, 7 years of experience) with access to patients’ clinical and imaging information. Both visual assessment and semiquantitative analyses were performed.

On visual assessment, a lesion was considered positive if focal ¹⁸F-FAPI-04 or ¹⁸F-FDG uptake exceeded that of adjacent normal tissue after excluding physiological uptake and benign findings.

Semiquantitative analyses were performed on the Syngo 6.5 (2012A) workstation. On axial PET and/or CT images, volumetric regions of interest (ROI) encompassing the entire lesion were delineated. ROIs were manually reviewed and adjusted to exclude physiological or inflammatory uptake. The software automatically generated quantitative parameters, including metabolic tumor volume (MTV) and total lesion glycolysis (TLG) for FDG PET, and FAPI-avid tumor volume (FTV) and total lesion FAP expression (TLF) for FAPI PET. CT served as the anatomic reference to ensure that standardized uptake value (SUV) measurements corresponded to the same lesion on both datasets. For each lesion (excluding PMs), we recorded lesion counts and maximum SUV (SUVmax), as well as MTV, TLG, FTV, and TLF[9]. The tumor-to-background ratio (TBR) was calculated as lesion SUVmax divided by background SUVmean. Background regions were defined as follows: Mediastinal blood pool for metastatic lymph nodes and pleural/peritoneal lesions; normal liver parenchyma for hepatic metastases; and the L5 vertebral body for bone metastases (BMs)[9]. To ensure standardization, identical background regions were applied for both ¹⁸F-FAPI-04 and ¹⁸F-FDG PET/CT in each lesion category. To assess whether inter-tracer variability in background uptake could influence TBR comparisons, background SUVmean values for each region were compared between the two tracers using paired t-tests.

For nodular PMs, lesion counts and metabolic parameters were recorded as the mean values across all lesions; when > 5 lesions were present, mean values were calculated from the five lesions with the highest uptake as a representative average. For diffuse PMs, lesion metabolic parameters were recorded. The extent of peritoneal disease was evaluated using the peritoneal cancer index (PCI) derived from PET/CT images; PCI quantifies tumor burden across Sugarbaker’s 13 abdominopelvic regions according to lesion size[3].

Reference standard

Histopathological findings from surgery or biopsy served as the reference standard. For lesions without pathological confirmation, final diagnosis was determined based on laboratory results, multimodal imaging [contrast-enhanced CT, PET/CT, or magnetic resonance (imaging)] follow-up, and comprehensive clinical assessment. All patients were required to undergo a follow-up period of at least three months. When pathological results and imaging follow-up were discordant, all available data (including additional clinical and imaging follow-up information) were jointly reviewed by two nuclear medicine physicians and one radiologist, with a final classification reached by consensus. If consensus could not be achieved, pathological results served as the reference standard.

Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics (version 26.0). Normally distributed continuous variables are presented as mean ± SD, whereas non-normally distributed variables are presented as median (Q1, Q3). Categorical variables are reported as frequencies (percentages). For paired comparisons between ¹⁸F-FAPI-04 PET/CT and ¹⁸F-FDG PET/CT, lesion-based quantitative parameters (SUVmax, TBR, MTV, TLG, FTV, and TLF) were compared using the Wilcoxon signed-rank test. The Mann-Whitney U test was applied when comparisons involved independent groups, as appropriate. Paired categorical outcomes were compared using the McNemar test. Interobserver agreement for lesion interpretation on ¹⁸F-FAPI-04 PET/CT and ¹⁸F-FDG PET/CT was assessed using Gwet’s agreement coefficient, interpreted as follows: ≤ 0.20, poor; 0.21-0.40, fair; 0.41-0.60, moderate; 0.61-0.80, substantial; and 0.81-1.00, almost perfect agreement. A two-sided P value < 0.05 was considered statistically significant.

RESULTS

Patient characteristics

Between February 2023 and March 2024, 31 patients with GC (either newly diagnosed or with suspected metastasis during postoperative surveillance) were scheduled to undergo paired ¹⁸F-FAPI-04 and ¹⁸F-FDG PET/CT. Ten patients were excluded for the following reasons: ¹⁸F-FAPI-04 PET/CT performed without ¹⁸F-FDG PET/CT within 7 days (n = 2), an inter-scan interval ≥ 7 days (n = 1), concomitant malignancy (n = 1), incomplete clinical or follow-up data (n = 5), and severe hepatic/renal dysfunction (n = 1). Ultimately, 21 patients were enrolled, including 11 newly diagnosed GC patients for initial staging and 10 postoperative patients with suspected recurrence or metastasis for restaging. The patient selection flowchart is shown in Figure 1. Among the 21 included patients, 18 (85.7%) were older than 60 years. Newly diagnosed GC accounted for 52.4% (11/21), and postoperative suspected local recurrence or metastasis accounted for 47.6% (10/21); notably, only 1 patient (4.8%, 1/21) had suspected local recurrence. In addition, 8 patients (38.1%) had lymph node metastases (LNMs), 2 (9.5%) had BMs, 1 (4.8%) had liver metastases, and 6 (28.6%) had PMs (Table 1).

Figure 1 Study population flowchart. FAPI: Fibroblast activation protein inhibitor; FDG: Fluorodeoxyglucose; PET/CT: Positron emission tomography/computed tomography.

Table 1 Baseline patient characteristics of the study patients.

Description of patients	n (%)
Age
> 60	18 (85.7)
≤ 60	3 (14.3)
Gender
Female	4 (19.0)
male	17 (81.0)
Primary tumors	11 (52.4)
LR or metastases	10 (47.6)
LNM	8 (38.1)
BM	2 (9.5)
LM	1 (4.8)
PM	6 (28.6)

LR: Local recurrences; LNM: Lymph node metastases; BM: Bone metastases; LM: Liver metastases; PM: Peritoneal metastases.

Gwet’s agreement coefficients showed substantial interobserver agreement for both ¹⁸F-FAPI-04 PET/CT (0.77) and ¹⁸F-FDG PET/CT (0.70). Inter-tracer comparison of background activity showed no statistically significant differences between ¹⁸F-FAPI-04 and ¹⁸F-FDG in any of the three reference regions: Mediastinal blood pool (P = 0.701), normal liver (P = 0.068), or L5 vertebral body (P = 0.134). These findings demonstrate that inter-tracer background variability does not substantially influence TBR comparisons.

Comparison of ¹⁸F-FAPI-04 and ¹⁸F-FDG uptake in primary tumors

Among the 11 patients with pathologically confirmed, newly diagnosed GC, primary tumors demonstrated significantly higher uptake and higher lesion-to-background contrast on ¹⁸F-FAPI-04 than on ¹⁸F-FDG (Table 2). SUVmax was higher with ¹⁸F-FAPI-04 than with ¹⁸F-FDG [11.6 (10.5-19.3) vs 6.5 (5.0-8.1), P = 0.036], and TBR was also higher [11.5 (10.5-17.4) vs 5.1 (4.6-7.4), P = 0.009]. Tracer-specific total-lesion metrics differed, with higher FAPI-derived TLF than FDG-derived TLG [260.5 (131.6-363.1) vs 57.9 (30.9-126.8), P = 0.036]. Although FTV tended to be higher than MTV [41.2 (17.9-69.9) vs 18.4 (8.6-27.9), P = 0.088], the difference did not reach statistical significance (Table 2; Figure 2).

Figure 2 A 76-year-old male underwent positron emission tomography/computed tomography to evaluate a gastric mass previously detected through Esophagogastroduodenoscopy. A: ¹⁸F-fibroblast activation protein inhibitor-04 positron emission tomography/computed tomography (PET/CT) images show one extremely strong uptake lesion (maximum standardized uptake value = 19.88) at the lesser curvature (solid line arrow), which were suspected of gastric cancer; B: ¹⁸F-fluorodeoxyglucose PET/CT images show mild tracer uptake (maximum standardized uptake value = 5.50) at the lesser curvature lesion (dotted line arrow). ¹⁸F-fibroblast activation protein inhibitor-04 PET/CT offered superior imaging contrast to ¹⁸F-fluorodeoxyglucose PET/CT, as evidenced by a significantly higher tumor-to-background ratio (19.88 vs 3.07) and better-defined tumor margins; C: The pathological findings derived from surgical resection are indicative of mucinous adenocarcinoma. HE: Hematoxylin-eosin; FAP: Fibroblast activation protein.

Table 2 Comparison of ¹⁸F-fibroblast activation protein inhibitor-04 and ¹⁸F-fluorodeoxyglucos uptake in various lesions.

Descriptions of lesions	18F-FAPI-04 PET/CT				18F-FDG PET/CT				P value FAPI/FDG
	SUVmax	FTV (cm3)	TLF (SUVbw × cm3)	TBR	SUVmax	MTV (cm3)	TLG (SUVbw × cm3)	TBR	SUVmax	TBR	F/MTV	TLF/G
Primary GC	11.6 (10.5-19.3)	41.2 (17.9-69.9)	260.5 (131.6-363.1)	11.5 (10.5-17.4)	6.5 (5.0-8.1)	18.4 (8.6-27.9)	57.9 (30.9-126.8)	5.1 (4.6-7.4)	0.036	0.009	0.088	0.036
LNM	6.6 (4.4-10.6)	3.1 (0.8-3.4)	8.3 (2.6-20.9)	5.4 (4.2-9.2)	6.5 (3.8-10.5)	1.9 (0.5-3.3)	6.3 (2.1-16.2)	6.4 (4.5-9.6)	0.703	0.65	0.321	0.654
BM	11.1 (9.0-16.6)	6.4 (3.2-14.8)	27.7 (12.5-84.2)	10.4 (8.4-16.0)	5.5 (4.5-6.7)	2.4 (0.9-13.1)	9.5 (2.8-52.5)	5.6 (4.0-7.0)	< 0.001	< 0.001	0.128	0.045
PM	7.1 (4.7-10.0)	13.2 (3.4-25.5)	39.3 (16.2-76.8)	7.4 (5.8-9.7)	3.2 (0.0-3.9)	1.2 (0.0-1.8)	3.3 (0.0-5.3)	2.4 (0.0-3.9)	0.002	0.002	0.003	0.004
LM	8.4 (7.4-9.3)	15.2 (10.3-20.1)	67.6 (43.7-91.5)	9.8 (8.7-10.9)	8.2 (8.1-8.2)	8.0 (6.3-9.7)	35.3 (27.9-42.8)	7.4 (6.9-7.8)	N/A	N/A	N/A	N/A
LR	3.9 (NA)	7.8 (NA)	24.7 (NA)	8.1 (NA)	2.9 (NA)	1.2 (NA)	3.2 (NA)	1.2 (NA)	N/A	N/A	N/A	N/A

¹⁸F-FAPI-04: ¹⁸F-fibroblast activation protein inhibitor-04; PET/CT: Positron emission tomography/computed tomography; ¹⁸F-FDG: ¹⁸F-fluorodeoxyglucose; FAPI: Fibroblast activation protein inhibitor; FDG: Fluorodeoxyglucose; SUVmax: Maximum standardized uptake value; FTV: Fibroblast activation protein inhibitor-avid tumor volume; TLF: Total lesion fibroblast activation protein expression; SUVbw: Standardized uptake value normalized to body weight; TBR: Tumor-to-background ratio; MTV: Fluorodeoxyglucose-derived metabolic tumor volume; TLG: Total lesion glycolysis; F/MTV: Fibroblast activation protein inhibitor-avid tumor volume/fluorodeoxyglucose-derived metabolic tumor volume; TLF/G: Total lesion fibroblast activation protein expression/total lesion glycolysis; GC: Gastric cancer; LR: Local recurrences; LNM: Lymph node metastases; BM: Bone metastases; LM: Liver metastases; PM: Peritoneal metastases; NA: Not available (data could not be calculated due to only 1 lesion in the local recurrences group); N/A: Not applicable (statistical comparisons were not performed for lesion types with fewer than 3 lesions i.e., liver metastases and local recurrences subgroups).

Comparison of ¹⁸F-FAPI-04 FAPI and ¹⁸F-FDG uptake in LNMs

In patients with LNMs confirmed by pathology and/or imaging follow-up, lesion-based analyses showed no significant differences between tracers (Table 2). Quantitative parameters were comparable, including SUVmax [6.6 (4.4-10.6) vs 6.5 (3.8-10.5), P = 0.703], FTV vs MTV [3.1 (0.8-3.4) vs. 1.9 (0.5-3.3), P = 0.321], and TLF vs TLG [8.3 (2.6-20.9) vs 6.3 (2.1-16.2), P = 0.654]. TBR likewise did not differ significantly between ¹⁸F-FAPI-04 and ¹⁸F-FDG [5.4 (4.2-9.2) vs 6.4 (4.5-9.6), P = 0.65] (Table 2; Figure 3).

Figure 3 A 79-year-old male with gastric discomfort underwent both ¹⁸F-fibroblast activation protein inhibitor-04 and ¹⁸F-fluorodeoxyglucos positron emission tomography/computed tomography imaging for initial assessment. A: ¹⁸F-fibroblast activation protein inhibitor-04 positron emission tomography/computed tomography (PET/CT) images show mildly increased fluorodeoxyglucose (FDG) uptake (maximum standardized uptake value = 3.1) in the lymph nodes adjacent to the gastric antrum (solid line arrow); B: ¹⁸F-FDG PET/CT images also show mild tracer uptake (maximum standardized uptake value = 3.2) in the lymph nodes adjacent to the gastric antrum (dotted line arrow). Compared with ¹⁸F-FDG PET/CT, ¹⁸F-fibroblast activation protein inhibitor-04 PET/CT did not demonstrate a high tumor-to-background ratio (2.30 vs 2.39) and clearer tumor contour; C: The pathological findings derived from lymph node resection indicated the presence of metastatic disease. HE: Hematoxylin-eosin; FAP: Fibroblast activation protein.

Comparison of ¹⁸F-FAPI-04 and ¹⁸F-FDG uptake in BMs

Imaging follow-up identified BMs in two patients. Compared with ¹⁸F-FDG, ¹⁸F-FAPI-04 demonstrated significantly higher SUVmax [11.1 (9.0-16.6) vs 5.5 (4.5-6.7), P < 0.001] and higher TBR [10.4 (8.4-16.0) vs 5.6 (4.0-7.0), P < 0.001] (Table 2). Tracer-specific total-lesion metrics also differed [TLF vs TLG: 27.7 (12.5-84.2) vs 9.5 (2.8-52.5), P = 0.045]. In contrast, FTV vs MTV did not differ significantly [6.4 (3.2-14.8) vs 2.4 (0.9-13.1), P = 0.128] (Table 2). However, given the limited numbers for BMs (n = 2), these findings should be interpreted as exploratory and hypothesis-generating.

Comparison of ¹⁸F-FAPI-04 and ¹⁸F-FDG uptake in PMs

PMs were identified in six patients by pathology and/or imaging follow-up. For nodular PMs, ¹⁸F-FAPI-04 demonstrated significantly higher uptake, contrast, and burden-related parameters than ¹⁸F-FDG, including SUVmax [7.1 (4.7-10.0) vs 3.2 (0.0-3.9), P = 0.002], TBR [7.4 (5.8-9.7) vs 2.4 (0.0-3.9), P = 0.002], FTV vs MTV [13.2 (3.4-25.5) vs 1.2 (0.0-1.8), P = 0.003], and TLF vs TLG [39.3 (16.2-76.8) vs 3.3 (0.0-5.3), P = 0.004] (Table 2; Figure 4). For diffuse PMs, PCI and SUVmax were further compared. ¹⁸F-FAPI-04 yielded a significantly higher PCI score [14 (6-25) vs 2.5 (2.0-3.0), P = 0.012] and higher SUVmax [8.0 (5.7-13.0) vs 4.7 (3.5-5.8), P = 0.001] than ¹⁸F-FDG (Table 3).

Figure 4 A 64-year-old male with a history of gastric cancer surgery underwent positron emission tomography/computed tomography to monitor for recurrence and metastases. A: ¹⁸F-fibroblast activation protein inhibitor-04 positron emission tomography/computed tomography images displayed intensely diffuse uptake (maximum standardized uptake value = 11.8) in the peritoneal metastases (solid line arrow); B: ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography images displayed no uptake in peritoneal metastatic lesions (dotted line arrow). Additionally, fibroblast activation protein inhibitor demonstrated increased uptake in more lymph node and muscular lesions compared to fluorodeoxyglucose.

Table 3 Comparison of ¹⁸F-fibroblast activation protein inhibitor-04 and ¹⁸F-fluorodeoxyglucos in diffuse peritoneal metastatic lesions.

Characteristics	FAPI		FDG		P value
	Median range		Median range
PCI	14	6-25	2.5	2-3	0.012
SUVmax	8.0	5.7-13.0	4.7	3.5-5.8	0.001
Characteristics	Primary tumors	LNM	BM	PM	LM
18F-FAPI-04	19	32	65	30	3
18F-FDG	17	25	36	6	2

PCI: Peritoneal cancer index; LR: Local recurrences; LNM: Lymph node metastases; BM: Bone metastases; LM: Liver metastases; PM: Peritoneal metastase; FAPI: Fibroblast activation protein inhibitor; FDG: Fluorodeoxyglucos; SUVmax: Maximum standardized uptake value; ¹⁸F-FAPI-04: ¹⁸F-fibroblast activation protein inhibitor-04; ¹⁸F-FDG: ¹⁸F-fluorodeoxyglucos.

Diagnostic performance of ¹⁸F-FAPI-04 PET/CT and ¹⁸F-FDG PET/CT

In the 11 patients with primary GC, ¹⁸F-FAPI-04 PET/CT detected all primary tumors, yielding a sensitivity of 100%, whereas ¹⁸F-FDG PET/CT achieved a sensitivity of 90.9% (Table 4). For metastatic disease, ¹⁸F-FAPI-04 PET/CT demonstrated sensitivities of 100% for both LNMs and PMs, whereas ¹⁸F-FDG PET/CT showed sensitivities of 75% and 50%, respectively. Given the study design and limited sample size (including only one true positive per tracer for BM, LM, and LR), robust calculation of false-positive rates or overall diagnostic accuracy was not feasible. Finally, lesion-count–based visualization comparing detection yields between the two tracers is shown in Figure 5. Across a total of 237 lesions, the Figure 5 indicates that compared with ¹⁸F-FDG PET/CT, ¹⁸F-FAPI-04 PET/CT identified more lesions, particularly for BMs, PMs. This observation is presented as descriptive only. No formal statistical testing was applied to these comparisons.

Figure 5 Compare the performance of ¹⁸F-fluorodeoxyglucose and ¹⁸F-fibroblast activation protein inhibitor-04 positron emission tomography in detecting primary tumors, lymph nodes, bone and peritoneal metastases, local recurrences. Data are presented descriptively without statistical hypothesis testing. No P values are shown. LNM: Lymph node metastases; Mets: Metastases; Recs: Recurrences; PET/CT: Positron emission tomography/computed tomography; ¹⁸F-FAPI: ¹⁸F-fibroblast activation protein inhibitor; ¹⁸F-FDG: ¹⁸F-fluorodeoxyglucose.

Table 4 Diagnostic performance of ¹⁸F-fibroblast activation protein inhibitor-04 and ¹⁸F-fluorodeoxyglucos uptake in various lesions.

Lesions	Tracer	No. of TP	No. of TN	No. of FN	Sen (%)	95%CI	NPV (%)
Primary tumors	18F-FAPI-04	11	0	0	100%	0.717-1.0	N/A
	18F-FDG	10	0	1	90.9%	0.623-0.984	N/A
LNM	18F-FAPI-04	8	1	0	100%	0.678-1.0	100%
	18F-FDG	6	1	2	75%	0.409-0.928	33.3%
BM	18F-FAPI-04	1	0	1	N/A	N/A	N/A
	18F-FDG	1	0	1	N/A	N/A	N/A
PM	18F-FAPI-04	6	0	0	100%	0.610-1.0	N/A
	18F-FDG	3	0	3	50%	0.188-0.812	N/A
LM	18F-FAPI-04	1	0	1	N/A	N/A	N/A
	18F-FDG	1	0	1	N/A	N/A	N/A
LR	18F-FAPI-04	1	0	0	N/A	N/A	N/A
	18F-FDG	1	0	0	N/A	N/A	N/A

¹⁸F-FAPI-04: ¹⁸F-fibroblast activation protein inhibitor-04; ¹⁸F-FDG: ¹⁸F-fluorodeoxyglucos; CI: Confidence interval; TP: True positive; TN: True negative; FN: False negative; NPV: Negative-predictive value; Sen: Sensitivity; LR: Local recurrences; LNM: Lymph node metastases; BM: Bone metastases; LM: Liver metastases; PM: Peritoneal metastases; N/A: Not applicable. Sensitivity not reported because only 1 true positive case was identified for this lesion type; any calculated sensitivity (50% or 100%) would be statistically unstable and clinically uninformative.

DISCUSSION

Accurate diagnosis and staging are essential for precision management and prognostic stratification in patients with GC. ⁶⁸Ga-labeled FAPI tracers target the tumor microenvironment and have demonstrated promising clinical utility in GC for tumor detection, initial staging, and evaluation of recurrence[5,6,10,11]. Among available derivatives, ⁶⁸Ga-FAPI-04 is considered an attractive probe for FAP-targeted imaging because of its high target affinity and favorable pharmacokinetic profile[12]. Nevertheless, wider clinical implementation of ⁶⁸Ga-labeled FAPI may be constrained by the short physical half-life of ⁶⁸Ga (67.8 minutes), higher costs, and reliance on a ⁶⁸Ge/⁶⁸Ga generator. In comparison,¹⁸F-labeled FAPI-04 offers several potential advantages over ⁶⁸Ga-FAPI-04. The longer half-life of ¹⁸F (109.8 minutes) allows for extended imaging windows and flexible scheduling, while the shorter positron range of ¹⁸F yields inherently better spatial resolution and image quality. Additionally, ¹⁸F can be produced in large quantities via cyclotron-based automated synthesis and distributed to sites lacking an on-site generator, substantially improving production efficiency and clinical accessibility. ¹⁸F also remains the most widely used PET radionuclide[13]. These properties make ¹⁸F-FAPI-04 a more feasible and scalable alternative to ⁶⁸Ga-FAPI-04 for routine clinical use. In this context, the present prospective pilot study directly compared ¹⁸F-FAPI-04 PET/CT with ¹⁸F-FDG PET/CT in patients with newly diagnosed GC and in postoperative patients with suspected recurrence or metastasis. Overall, we observed trends of higher tracer uptake (SUVmax), higher TBR, and higher tracer-specific tumor-burden metrics, particularly total lesion FAP expression (TLF), with ¹⁸F-FAPI-04 PET/CT compared to ¹⁸F-FDG PET/CT in primary tumors, PMs, and BMs (all P < 0.05). Importantly, our findings extend the evidence supporting the clinical application of ¹⁸F-labeled FAPI by establishing its feasibility and diagnostic potential in a head-to-head comparison with ¹⁸F-FDG. The volumetric parameters reported in this study (FTV/TLF vs MTV/TLG) are tracer-specific: ¹⁸F-FAPI-04-derived TLF reflects the total burden of FAP expression in the tumor microenvironment, whereas ¹⁸F-FDG-derived TLG reflects glucose metabolic activity. Our interpretation of higher TLF should be understood within this context. Higher TLF may reflect greater tumor burden, higher FAP expression density, or both, and should not be directly interpreted as evidence that ¹⁸F-FAPI-04 is volumetrically superior to ¹⁸F-FDG. Therefore, these tracer-specific metrics cannot be directly equated across tracers.

Prior studies have suggested that FAPI-04 PET/CT may outperform FDG PET/CT for detecting primary GC[14,15]. In a prospective comparison across gastrointestinal cancers, Yang et al[15] found that ¹⁸F-FAPI-04 detected primary tumors more often than ¹⁸F-FDG in various gastrointestinal system cancers and found that FAPI had significant advantages in detecting primary tumors (97.7% vs 72.7%). Subsequently, Lv et al[16] subsequently examined GC specifically and reported that ¹⁸F-FAPI-04 achieved 100% sensitivity for primary or recurrent tumors, with particularly strong performance in low or no FDG uptake subtypes such as signet-ring cell carcinoma, supporting its role as a precise staging tool for GC. Using the related tracer ¹⁸F-FAPI-74 in patients with gastric, liver, and pancreatic cancers, a further prospective study likewise found FAPI superior to ¹⁸F-FDG in the detection of primary tumors [88% (22/25) vs 60% (15/25), P = 0.016][7]. In our cohort, ¹⁸F-FAPI-04 PET/CT detected all primary tumors (100%, 11/11) and achieved a higher TBR than ¹⁸F-FDG PET/CT (11.5 vs 5.1, P = 0.009), whereas ¹⁸F-FDG PET/CT detected 10 of 11 tumors (90.9%). The single ¹⁸F-FDG-negative primary tumor was pathologically confirmed as poorly differentiated adenocarcinoma with a signet-ring cell component, consistent with previous findings[16,17]. Mucinous adenocarcinoma and signet-ring cell carcinoma are recognized to exhibit relatively low glucose-transporter expression, which may reduce the sensitivity of ¹⁸F-FDG PET/CT in these subtypes. In addition, physiological gastric wall uptake can further complicate FDG-based evaluation of primary lesions. In contrast, FAPI-04 typically shows low background activity in the abdominopelvic region, which may facilitate delineation of primary GC[11]. Moreover, factors reported to influence tumor affinity of ⁶⁸Ga-FAPI-04 include depth of invasion and tumor size rather than differentiation grade, histologic subtype, or Lauren classification[18]. Consistent with this, although primary-tumor detection did not differ substantially between tracers in our cohort, primary tumors exhibited markedly higher uptake on ¹⁸F-FAPI-04 than on ¹⁸F-FDG.

Of note, false-positive uptake of FAPI was mainly observed in inflammatory or fibrotic lesions, indicating that its specificity remains a limitation. However, the absolute difference in primary-tumor detection between tracers was modest in this study, which may be related to: (1) The limited sample size and relatively homogeneous histology (predominantly adenocarcinoma, with only one case containing a signet-ring component); and (2) The predominance of advanced-stage disease, with larger tumor burden that may increase FDG avidity and thereby reduce the apparent separation between tracers.

Whether ¹⁸F-FAPI-04 provides a consistent advantage for detecting LNMs remains controversial and requires validation in larger cohorts. In our lesion-based analysis, metastatic lymph nodes showed no significant differences in uptake or derived parameters between tracers (P > 0.05), and TBR was also comparable, consistent with reports by Kuten et al[14] and Miao et al[18]. By contrast, Pang et al[19] reported superior overall performance of ⁶⁸Ga-FAPI relative to ¹⁸F-FDG for LNM detection, although no significant differences were observed for mediastinal nodal metastases. One plausible explanation is the biological complexity of the nodal metastatic microenvironment, which involves dynamic interactions among infiltrating tumor cells, immune responses, and fibrosis-related processes[20].

Recent reports have suggested that ¹⁸F-FAPI-04 may achieve higher sensitivity than ¹⁸F-FDG for LNM detection and may demonstrate higher nodal uptake on FAPI imaging[16,21], potentially driven by improved uptake and lesion-to-background contrast in metastatic nodes[17]. For nodal disease, Yang et al[15] reported a sensitivity of 91.89% for ¹⁸F-FAPI-04 PET/CT achieved a sensitivity of 91.89% for LNM in gastrointestinal cancers vs 79.72% for ¹⁸F-FDG. and Lv et al[16] similarly reported higher nodal sensitivity in GC (88.89% vs 38.89%). Although our study did not show statistically significant quantitative differences for LNMs, possibly due to limited sample size and pooling of preoperative and postoperative patients without stratified analyses, our lesion-count-based visualization suggested that ¹⁸F-FAPI-04 PET/CT has the potential to identify more nodal lesions, or a broader extent of nodal disease, than ¹⁸F-FDG, consistent with observations by Qin et al[6].

PM is a common dissemination route in GC, and approximately 22.8% of patients may still develop PM despite multimodal therapy[22]. The presence of PM is associated with poor prognosis and represents one of the most frequent causes of death in GC[23]. In this study, ¹⁸F-FAPI-04 PET/CT was superior to ¹⁸F-FDG PET/CT for detecting PM, with higher uptake (SUVmax), higher contrast (TBR), larger volumetric parameters, and broader disease extent as reflected by higher PCI, consistent with previous reports[15-16]. In our study, PCI was derived from PET/CT imaging rather than from surgical exploration, and these imaging-derived PCI scores were not systematically validated against operative findings in the entire cohort. Only one patient underwent concurrent surgical exploration, too few for formal statistical validation. Imaging-derived PCI may diverge from operative findings, either by missing small-volume disease or by overinterpreting physiologic or inflammatory uptake. Therefore, the reported PCI scores may underestimate or overestimate the true peritoneal disease burden, and caution is warranted when comparing them with studies using surgically derived PCI.

For distant metastatic disease, such as BM and hepatic metastases, both tracers detected BMs in two patients, and hepatic metastasis in one patient in our cohort. However, compared with ¹⁸F-FDG, metastatic lesions demonstrated higher uptake and higher TBR on ¹⁸F-FAPI-04 PET/CT, and FAPI-04 appeared to detect more lesions, particularly small lesions, consistent with prior studies[15,24,25]. Potential explanations include low physiological uptake in the gastrointestinal tract, relatively low hepatic background activity on FAPI imaging, and high FAPI avidity in osseous metastatic lesions[26]. Nevertheless, these findings require confirmation in larger studies.

In summary, ¹⁸F-FAPI-04 PET/CT improves lesion conspicuity and reduces the risk of missed disease in GC, particularly for LNM and PM, although these imaging findings remain observational and hypothesis-generating. This advantage most likely arises from tumor–stroma interactions. CAFs, the most abundant stromal component of the GC tumor microenvironment, strongly express the FAPI target FAP[19], though the level of FAP expression varies between tumors, which may underlie the inconsistent performance of FAPI-based tracers across studies[27]. Accordingly, FAPI imaging detects metastatic lesions in GC subtypes with low FDG avidity, including signet-ring cell and poorly differentiated adenocarcinomas[14]. CAF behavior is itself heterogeneous across metastatic sites, with variable FAP expression and matrix remodeling capacity[18]. In PM, CAF recruitment and activation upregulate FAP, and the minimal physiological uptake of FAPI in normal tissues yields a high target-to-background ratio, enabling clear delineation of diffuse carcinomatosis even in FDG-negative patients[14,18]. In LNM, involved lymph nodes are enriched with FAP-positive CAFs derived from both tumor-induced stromal activation and the transformation of resident fibroblasts[19]. Beyond their impact on diagnostic imaging, CAFs also help shape the tumor immune microenvironment. Narote et al[28] reported that CAFs construct an immunosuppressive microenvironment through excessive collagen deposition, secretion of transforming growth factor β/interleukin-6/interleukin-10, lactate- and hypoxia-driven metabolic reprogramming, and modulation of immune cell migration and function. These processes may act in concert with extracellular matrix collagen-binding receptors (leukocyte-associated immunoglobulin-like receptor 1, osteoclast-associated receptor, discoidin domain receptors 1 or 2) and immune checkpoints (programmed cell death protein 1/programmed cell death protein 1 ligand, cytotoxic T-lymphocyte-associated protein 4, lymphocyte activation gene-3, T-cell immunoglobulin and mucin domain-3, T cell immunoreceptor with immunoglobulin and ITIM domain), thereby limiting the efficacy of immune checkpoint inhibitors. Although mechanistically distinct from FAP-targeted imaging, evidence from other disease models supports a broader principle that molecular targeting of pathogenic drivers - whether FAP in cancer or specific autoantibodies in neurological disorders - can yield both diagnostic and therapeutic benefits[29-32].

Collectively, our results position ¹⁸F-FAPI-04 PET/CT as a promising tool for improving staging accuracy and reducing missed metastases in GC. However, given the pilot design and small cohort, these findings warrant validation in multicenter studies with larger cohorts. Integrating radiomic or functional imaging phenotypes with multi-omics data - genomic, transcriptomic, proteomic, and epigenomic profiling - could further elucidate the functional interactions and tumor-specific vulnerabilities underlying the heterogeneous tracer uptake observed across metastatic sites[33].

This study has several limitations. First, we pooled newly diagnosed and postoperative patients without statistical stratification because the cohort was very small (newly diagnosed, n = 10; postoperative, n = 11), with particularly few cases of bone metastasis (n = 2), liver metastasis (n = 1), and local recurrence (n = 1). These two populations differ in disease biology and clinical context (preoperative staging vs postoperative surveillance), so the statistically significant differences we observed (in SUVmax, TBR, and total-lesion metrics) are exploratory and hypothesis-generating. These preliminary findings do not constitute conclusive evidence of tracer performance, and larger studies with prespecified subgroup analyses will be needed to validate our observations. Second, we did not perform clustering correction in the lesion-based analyses (237 lesions) to account for the correlation among lesions within the same patient. Given the non-independence of multiple lesions from the same patient, the statistical significance of lesion-level findings may be inflated. Future studies should use multilevel models or cluster-robust standard errors for lesion-based analyses. Third, not all lesions were histopathologically confirmed; a proportion were adjudicated by imaging follow-up and clinical assessment. Although this approach is standard in clinical imaging research, it introduces verification bias and precludes robust estimation of specificity and overall diagnostic accuracy within the present cohort. Therefore, our analysis primarily addresses the comparative detection yield of GC-related lesions. These diagnostic performance estimates, especially specificity, should be interpreted cautiously, and prospective studies with systematic histopathological correlation are needed to confirm our findings.

CONCLUSION

In this pilot study, ¹⁸F-FAPI-04 PET/CT generally showed higher tracer uptake and lesion-to-background contrast than ¹⁸F-FDG PET/CT in primary GC, PMs, and BMs. It also tended to detect more metastatic lesions overall, with higher detection yields particularly for PM, suggesting its potential value for initial staging and postoperative assessment in GC. However, given the single-center design and limited sample size, these findings require confirmation in larger, multicenter studies.