Computed tomography-guided percutaneous biopsy for assessing tumor heterogeneity in neuroendocrine tumor metastases to the liver

Abstract

BACKGROUND

Gastroenteropancreatic neuroendocrine tumors (GEP-NETs) frequently metastasize to the liver, with heterogeneity in tumor grade impacting patient prognosis and treatment. The Ki-67 index, a key prognostic marker, often varies between primary and metastatic sites; however, routine liver biopsy remains controversial. Although percutaneous computed tomography-guided core needle biopsy (PCT-CNB) is safe and effective for focal lesions, its role in detecting intertumor grading discrepancies and survival implications in GEP-NETs is underexplored. Conflicting survival associations with grade shifts have been reported in previous studies. We hypothesized that PCT-CNB could identify clinically significant grading heterogeneity in liver metastases, correlating with survival outcomes, thereby refining risk stratification and therapeutic strategies.

AIM

To investigate intertumor grading heterogeneity in GEP-NET liver metastases via PCT-CNB.

METHODS

We retrospectively investigated 92 patients with liver metastases from GEP-NETs via PCT-CNB, 76 patient samples from the liver and primary sites, and 16 from the liver and secondary liver sites. Ki-67 immunohistochemistry was performed for tissue sampling, and grading classifications were determined. Intertumor grading classification heterogeneity and associated changes in patient survival outcomes were also evaluated.

RESULTS

No procedure-related mortality was recorded during or after biopsy. In 37/92 patients (40.2%), the grading classifications changed: The grading increased from G1 to G2 in 13 patients, from G1 to G3 in 2, and from G2 to G3 in 14; the grading decreased from G2 to G1 in 5 patients, from G3 to G1 in 1, and from G3 to G2 in 2. Patients with G1 or G2 disease had better progression-free survival and overall survival (OS) outcomes than those with G3 disease did (P = 0.001 and P < 0.001, respectively). The 5-year and 10-year OS rates for stable G2 patients were 67.5% and 26.0%, respectively, decreasing to 46.4% and 23.2%, respectively, among G2 patients whose grade increased (P = 0.016).

CONCLUSION

The PCT-CNB of liver metastases from GEP-NETs differed in grade between the liver tumor and primary site/secondary liver metastases. Additionally, when grading increased from G2, the OS rate significantly decreased.

Key Words: Image-guided core needle biopsy; Gastroenteropancreatic neuroendocrine tumors; Liver metastases; Survival; Tumor heterogeneity

Core Tip: This study demonstrates that percutaneous computed tomography-guided core needle biopsy effectively identifies clinically significant grading heterogeneity in liver metastases from gastroenteropancreatic neuroendocrine tumors. Notably, 40.2% of patients exhibited grade discrepancies between primary and metastatic sites, with upgraded tumors (e.g., G2 to G3) correlating with markedly reduced overall survival. These findings emphasize percutaneous computed tomography-guided core needle biopsy’s pivotal role in optimizing risk stratification, informing personalized therapeutic decisions (e.g., intensified surveillance or targeted therapies), and improving prognostic accuracy for gastroenteropancreatic neuroendocrine tumor patients, while confirming its safety and diagnostic reliability in clinical practice.

INTRODUCTION

Neuroendocrine tumors (NETs) are generated in the neuroendocrine cell system; they form organoid cell aggregations or are composed of disseminated cells in different organs[1,2]. Gastroenteropancreatic NETs (GEP-NETs) are the main NET subtypes and account for approximately 66% of all NETs[3]. In recent years, GEP-NET incidence rates have increased. From 1973-2012, according to National Cancer Institute data[4], GEP NET incidence rates increased 6.4-fold. To combat GEP-NET, Ki-67 is used as a cell proliferation nuclear marker, it is an important prognostic disease variable[5,6], and Ki-67 indices are included in World Health Organization (2010) grading classifications for GEP NETs[7].

Liver metastases are frequently observed in patients with GEP-NETs; approximately 65%-95% of patients will manifest these symptoms, even when primary tumors are small[8,9]. Also, liver metastases generate a strong prognostic impact in these patients[10,11], while tumor heterogeneity may significantly impact clinical disease management. Previous studies have considered Ki-67 staining heterogeneity between primaries and metastases[12-15]. Therefore, liver biopsies are required for accurate evaluations, and primary and metastatic tumor biopsies are recommended by a National Expert Canadian Group[16]. However, some studies have indicated that primary tumor biopsies are not necessary if the liver metastasis biopsy provides positive and comprehensive tumor information[17].

Percutaneous computed tomography (CT) guided core needle biopsies (PCT-CNB) or ultrasound (US)-guided core needle biopsies are widely used to assess diagnostic and prognostic markers and potential therapeutic targets in focal liver lesions[18,19]. While US is predominantly used for biopsies, CT is used for lesions that are not well-visualized by US; this may be due to small lesion size, altered liver morphology, isoechogenicity, body shape/build, or lesions contained at difficult anatomical locations such as the liver dome[20,21]. PCT-CNB is a safe procedure and generates adequate sample quality to study tumor heterogeneity. We used this technique to assess inter-tumor grading classification heterogeneity in GEP-NETs. Tumor heterogeneity was investigated: (1) Between the liver tumor and primary sites; and (2) Between the liver tumor and secondary liver metastases. Additionally, the effects of grade classification changes on patient survival outcomes were investigated.

MATERIALS AND METHODS

All methods in this study were performed in accordance with the relevant guidelines and regulations.

Study participants

This retrospective study (No. 1612167-18) included patients diagnosed with histologically confirmed GEP-NETs and radiologically confirmed liver metastases who underwent PCT-CNB between August 2014 and December 2021 at our institution. The inclusion criteria were as follows: (1) Histopathological confirmation of primary GEP-NETs; (2) ≥ 1 liver metastases verified by contrast-enhanced CT or magnetic resonance imaging (MRI); (3) Availability of paired biopsy samples from both liver metastases and either primary tumors (n = 76) or secondary intrahepatic metastases (n = 16); and (4) Complete clinical and follow-up data. The exclusion criteria were as follows: Severe coagulation disorders (international normalized radio > 1.5 or platelet count < 50 × 10³/μL), contraindications to CT-guided biopsy (e.g., uncorrectable respiratory motion or inaccessible lesion location), concurrent malignancies or life-threatening comorbidities, and prior liver-directed therapies (e.g., resection or ablation) targeting the biopsied lesions. Among the 212 screened patients, 92 met the criteria and were included in the final analysis.

Study protocol

An enhanced CT or MRI scan was required within 30 days before PCT-CNB. The protocol also outlined PCT-CNB procedures, postoperative complication assessments, Ki-67 immunohistochemistry, Ki-67 assessments, follow-up procedures, and overall survival (OS) and progression-free survival (PFS) outcome assessments.

The PCT-CNB procedure

Biopsies were performed via a 64-slice spiral CT scanner (250 mA, 120 kV, and 3 mm thick; Philips Healthcare, Andover, MA, United States). One of five radiologists with > 5 years of experience performed the biopsies. First, an abdominal CT scan (planning procedure) was performed when the patient was in an appropriate position. During procedures, unenhanced CT images were reviewed with previous diagnostic contrast-enhanced CT or MRI images to locate isoattenuating target lesions, identify cross-sectional levels, and ascertain lesional relationships with internal liver structures (landmarks). After the skin was marked, sterile draping, local anesthesia, and the incision were performed, and radiologists performed a coaxial technique. For this purpose, 18-gauge biopsy and 17-gauge introducer needles (SuperCore; Argon Medical Devices; Plano, TX, United States) were advanced to the biopsy target edge.

The previous diagnostic contrast images and immediate CT images were compared via a side-by-side approach to avoid damage to structures such as the gallbladder, duodenum, pancreas, stomach, spleen, and portal vein. Tissue samplings numbered 2-3. CT images were immediately gathered after biopsy to identify any procedural issues. An abdominal belt was worn to reduce hepatic hemorrhage. Afterward, patients were admitted and observed for 4 h, and those without complications were discharged.

Ki-67 immunohistochemistry

Ki-67 antibodies were used to assess the tumor proliferation index in liver and primary tumor paraffin block sections. A two-step immunohistochemistry protocol was implemented; from the paraffin blocks, 4 μm sections were cut and stained with a Ki-67 antibody (MIB-1, Dako Corporation, CA, United States). Normal tonsil tissue and a mouse IgG antibody were used as positive and negative controls, respectively. Antigen retrieval was performed in citrate buffer (pH = 6) for 3 minutes under pressure. An Envision + Dual Link Kit (Dako Corporation, CA, United States) was used for detection. Diaminobenzidine was used as a highly sensitive chromogenic substrate, and hematoxylin was used as a counterstain. Brown nuclear staining indicated Ki-67 positivity.

Ki-67 assessment

The Ki-67 index was assessed by two independent pathologists blinded to the patients’ clinical outcomes. Each pathologist manually counted 2000 tumor cells in the selected hot spots (areas of highest Ki-67 positivity) from camera-captured images. Discrepancies in grading classifications (e.g., G1 vs G2) were resolved through joint re-evaluation and consultation with a third senior pathologist with over 15 years of experience in neuroendocrine pathology. The final grading was determined with consensus among all three evaluators. Dark brown tumor nuclei indicated Ki-67 positivity; light brown nuclei and cytoplasmic staining were excluded. The Ki-67 index indicates the ratio of positive cells to total counted cells (2000 cells) in highlighted areas. Then, the tumors were classified as G1 (Ki-67: 0%-2%), G2 (Ki-67: 2%-20%), or G3 (Ki-67: > 20%) per the 2010 World Health Organization criteria[22].

Follow-up procedures

Follow-up intervals were dependent on surgery, systemic treatment, or nonsurgical liver-directed therapy. In addition to overseeing physical symptoms, follow-up programs included physical examinations, radiological imaging, and laboratory tests. Some patients underwent somatostatin receptor scintigraphy (Octreoscan) or Ga68 positron emission tomography-CT as part of the follow-up assessments. All assessment results were recorded, including patient survival status, disease recurrence, and progression follow-up times (the last imaging time). The follow-up deadline was March 2022.

Statistical analysis

All qualitative data are expressed as percentages, and quantitative data are expressed as medians (ranges). Kaplan-Meier curves were generated to compare OS and PFS outcomes between predefined subgroups: (1) Stable G2 (no grade change between primary and metastasis) and (2) Increased G2 (G2 to G3 upgrade in metastasis). Missing data (e.g., incomplete treatment records in 5 patients) were minimal (< 5% of the cohort) and handled via complete-case analysis, as no significant bias was detected between included and excluded cases. Statistical analyses were performed using commercially available software (SPSS 20, IBM, Chicago, IL, United States). A P value < 0.05 was considered to indicate significance.

RESULTS

Clinical characteristics and pathological features

Patient clinical characteristics and pathological features are summarized in Table 1, which highlights the predominance of primary pancreatic tumors (81.5%) and the distribution of biopsy methods across liver segments (Figure 1). Figure 1D and F illustrate the CT-guided coaxial needle placement and postprocedural evaluation, demonstrating the technical feasibility of targeting lesions adjacent to critical vascular structures. Percutaneous biopsy procedures were successfully performed in 92 patients with 97 liver tumors, with no treatment-related deaths or severe complications. Five patients experienced self-limited perihepatic hemorrhages; however, interventions were not needed. Additionally, no tumor seeding or diaphragmatic injury was observed. Patient clinical characteristics and pathological features are shown in Table 1.

Figure 1 Computed tomography-guided biopsy of a pancreatic neuroendocrine tumor liver metastasis. A: Axial contrast-enhanced computed tomography (CT) reveals an arterially enhancing 45 cm lesion in the right lobe of the liver (arrow); B: Axial contrast-enhanced CT reveals an artery nearby the lesion (arrow); C: Axial contrast-enhanced CT reveals a portal vein nearby the lesion (arrow); D: Procedural non-enhanced CT image at the same cross-sectional level reveals the lesion to be hypoattenuating to the liver; E: 17-gauge introducer needle (arrow) was placed in approximate location of the lesion based on the avoidance of damage to artery within the liver and portal vein; F: Post-procedural non-enhanced CT image reveals no complications.

Table 1 Patient clinical characteristics and pathological features.

Characteristics	Number (median or mean ± SD)
Patients	92
Median age (range), years	48 (27-77)
Male/female	48/44
Liver tumors (PCT-CNB)	97
Tumors size (cm), mean ± SD	3.9 ± 2.7
Liver segments
S1	0 (0)
S2	3 (3.3)
S3	6 (6.5)
S4	5 (5.4)
S5	12 (12.0)
S6	29 (30.4)
S7	20 (20.7)
S8	22 (21.7)
Grading classifications (PCT-CNB)
G1	14 (14.4)
G2	61 (62.9)
G3	22 (22.7)
Grading classifications (initial diagnosis)
G1	20 (21.7)
G2	59 (64.1)
G3	13 (14.1)
Primary tumor site
Stomach	3 (3.3)
Duodenum	3 (3.3)
Pancreas	75 (81.5)
Jejunum/ileum	4 (4.3)
Rectum	7 (7.6)
Tumor samples method (liver and primary sites)
PCT-CNB	97 (52.7)
PUS-CNB	10 (5.4)
EUS-FNA	12 (6.5)
Gastroscopic/colonoscopic biopsy	4 (2.2)
Surgery	61 (33.2)

All included patients met predefined inclusion criteria, with no significant differences in baseline characteristics between subgroups (all P > 0.05). PCT-CNB: Percutaneous computed tomography-guided core needle biopsy; PUS-CNB: Percutaneous ultrasound-guided core needle biopsy; EUS-FNA: Endoscopic ultrasonography guided fine needle aspiration.

Tumor grading (in 92 patients) revealed 14 G1, 61 G2, and 22 G3 grades. At initial diagnosis, 20 G1, 59 G2, and 13 G3 grades were recorded. The primary tumors were located in the duodenum (n = 3), jejunum/ileum (n = 4), stomach (n = 3), rectum (n = 7), or pancreas (n = 75). In addition to 97 PCT-CNB biopsies, other tumor sample methods at liver or primary sites included percutaneous US-guided core needle biopsy (percutaneous US-guided core needle biopsy) (n = 10), endoscopic US-guided fine needle aspiration (EUS-FNA) (n = 12), gastroscopic/colonoscopic biopsy (n = 4), and surgery (n = 61).

Ki-67 tumor heterogeneity

Following a consensus review, 37/92 patients (40.2%) exhibited grade changes between primary and metastatic sites: 13 were upgraded from G1 to G2, 2 from G1 to G3, and 14 from G2 to G3; 5 were downgraded from G2 to G1, 5 from G2 to G1, 1 from G3 to G1, and 2 from G3 to G2.

Ki-67 heterogeneity between the liver metastases and primary tumor

In the 76 patients with Ki-67 staining of liver metastases and primary tumors, the secondary diagnostic tumor grades were compared with the initial diagnoses. We identified 25 patients (32.9%) with increased grades; 13 increased from G1 to G2, 2 from G1 to G3, and 10 from G2 to G3. We identified 6 (7.9%) patients with decreased grades; 4 decreased from G2 to G1, 1 from G3 to G1, and 1 from G3 to G2. Additionally, 45 patients (59.2%) had tumors with stable grades: 5 with G1, 32 with G2, and 8 with G3 (Table 2).

Table 2 Grade changes between liver tumor (percutaneous computed tomography-guided core needle biopsy) and the primary tumor (n = 76).

Patients	Grade at initial diagnosis	Grade at second diagnosis
5	G1	G1
13	G1	G2
2	G1	G3
4	G2	G1
32	G2	G2
10	G2	G3
1	G3	G1
1	G3	G2
8	G3	G3

Ki-67 heterogeneity between liver metastases

In 16 patients with Ki-67 staining of the liver and secondary liver sites, the secondary diagnostic tumor grades were compared with the initial diagnoses. We identified six patients (37.5%) with changed classifications: Four patient classifications increased from G2 to G3, one decreased from G2 to G1, and one decreased from G3 to G2. Ten patients (62.5%) had tumors with stable grades: Eight with G2 and 2 with G3 (Table 3).

Table 3 Grade changes between liver tumor (percutaneous computed tomography-guided core needle biopsy) and the secondary liver tumor (percutaneous computed tomography-guided core needle biopsy or percutaneous ultrasound-guided core needle biopsy or surgery) (n = 16).

Patients	Grade at initial diagnosis	Grade at second diagnosis
0	G1	G1
0	G1	G2
0	G1	G3
1	G2	G1
8	G2	G2
4	G2	G3
0	G3	G1
1	G3	G2
2	G3	G3

Survival analysis

The median follow-up time for all patients was 44.0 months (range: 4.3-240.0). The OS rates at 5 and 10 years were 89.2% and 64.5% for G1 tumors, 59.4% and 24.0% for G2 tumors, and 10.1% and 0.0% for G3 tumors (P < 0.001) (Figure 2A), respectively. Patients were stratified into two subgroups according to grade changes in liver metastases compared with the initial diagnosis: (1) “Stable G2” (n = 32), defined as tumors retaining G2 classification in both primary and metastatic sites; and (2) “Increased G2” (n = 14), defined as tumors upgraded from G2 (primary) to G3 (metastatic). For stable G2 patients, the OS rates at 5 and 10 years were 67.5% and 26.0%, respectively, whereas these rates decreased to 46.4% and 23.2%, respectively, for increased G2 patients (P = 0.016) (Figure 2B). PFS rates at 5 and 10 years were 62.4% and 41.6%, 34.3% and 0%, and 0% for G1, G2, and G3, respectively (P = 0.001) (Figure 2C).

Figure 2 Survival outcomes of gastroenteropancreatic neuroendocrine tumors: analysis of overall survival and progression-free survival based on initial grades and G2 group variations. Numbers at risk indicate patients remaining in follow-up at each time point, calculated via reverse Kaplan-Meier method (G1: Ki-67 ≤ 2%; G2: 3%-20%; G3: > 20%). A: The overall survival (OS) of patients with gastroenteropancreatic neuroendocrine tumors (GEP-NETs) with different grades at initial diagnosis. Patients with G1 or G2 had better overall survival rates when compared with patients with G3 (P < 0.001); B: The OS of patients with GEP-NETs in stable G2 and increased G2 groups. “Stable G2” (n = 32): Liver metastases retained G2 classification. “Increased G2” (n = 14): Liver metastases upgraded from G2 (primary) to G3. The OS of stable G2 patients was longer than for increased G2 patients (P = 0.016); C: Progression-free survival in patients with GEP-NETs with different grades at initial diagnosis. Patients with G1 or G2 had a better progression-free survival when compared with patients with G3 grades (P = 0.001).

DISCUSSION

We used PCT-CNB for liver metastases in GEP-NETs to identify inter-tumor grading classification heterogeneity, resulting in three important observations. First, PCT-CNB was safely performed with a sufficient tumor sample for tumor grading. Second, PCT-CNB showed frequent grade differences between liver tumors and primary site/secondary liver metastases. Third, when the G2 grade increased, the OS rate decreased significantly.

PCT-CNB for liver tumors is a minimally invasive tool that supplements clinical disease management, including diagnostics, prognostic assessments (disease staging), and/or therapeutic decision-making[20,23]. The biopsy technique may be performed in a coaxial or noncoaxial manner; however, the former technique is safer than other biopsy techniques, as multiple needle passes can be made with a single pleural puncture, thereby reducing pneumothorax incidence rates. The coaxial approach is particularly advantageous, as a single hepatic puncture permits multiple cuttings, whereas a single needle technique may generate two or more hepatic punctures per procedure[24]. For NETs, EUS-FNA sampling cannot generate specimens for histology, whereas the FNA yield may be insufficient to determine Ki-67 indices. Therefore, EUS-FNA appears suboptimal for the pretreatment grading of NETs[25,26].

We used the coaxial technique in this study, and 2-3 specimens were acquired and reflected standard practices at our institution. Importantly, sufficient tumor sampling material was obtained for liver metastasis classification grading of GEP-NETs. Moreover, the percutaneous biopsy procedures were technically successful, with no treatment-related deaths or severe complications recorded.

Inter-tumor Ki-67 heterogeneity values were previously reported[12,14,15]. Dhall et al[12] identified differences between different tumor sites in ileal NET, where 3/60 (5%) patients showed Ki-67 heterogeneity between different primary tumor sections. Grillo et al[14] identified increased Ki-67 Labeling indices between primary tumor and metastatic sites in 19/49 (38.8%) patients with GEP-NETs, while a recent study[15] identified increased grading classifications in 35/103 (34.0%) cases. Similar to these data, we observed similar rates of change between liver tumors and primary site/secondary liver metastases in 37/92 (40.2%) patients.

Previously, associations between increased grades and patient survival were evaluated[2,14,15], with shifts to next grades mainly occurring from G1 to G2[14,15]. Keck et al[15] reported that patients with increased metastasis grades performed poorly; however, no significant differences were reported for PFS (P = 0.55) or OS (P = 0.32). This observation concurred with our data where grades in 13 patients increased from G1 to G2. However, comparisons were not made for patient survival between the increased G1 group and the stable G1 group as this latter group was had low numbers (n = 5). While academically interesting, these data exert no impact on NET patients and their clinical management, as patients with G1 and G2 tumors receive the same treatment[27]. Additionally, in 10 of our patients with GEP-NET, the Ki-67 index changed from G2 to G3 during the disease course. Twenty-three patients had stable grade G2 tumors. OS rates at 5 and 10 years were 67.5% and 26.0% for the stable G2 group, but this decreased to 46.4% and 23.2% for the increased G2 group (P = 0.016). This finding agreed with previous reports[2,15,28] showing that increased grades were associated with decreased OS.

It is unclear why Ki-67 indices are often heterogeneous in GEP-NETs. Potential mechanisms include clonal evolution favoring proliferative subclones, which are linked to chromosomal instability in pancreatic NETs[29,30]. Treatment-induced selection of resistant clones may drive grade escalation, as observed in therapy-associated dedifferentiation[28]. Spatiotemporal biopsy disparities and hypoxia-inducible factor pathway activation in metastases likely contribute to phenotypic divergence[30]. It was previously hypothesized that this heterogeneity could be attributable to genetic variations in NETs, with similar explanations for other solid cancers. Another group theorized that Ki-67 index changes potentially result from therapy resistance and treatment effects[1,2]. Further studies are warranted to explore the underlying mechanisms involved.

G1 to G2 tumor changes in patients are considered well-differentiated; therefore, treatments do not necessarily change under existing clinical management approaches[15]. However, well-differentiated G2 GEP-NETs may progress to G3 tumors, which are morphologically differentiated and may require other therapeutic strategies[31]. Several strategies are available for the clinical management of liver metastases in GEP-NETs[17] ranging from surgery to ablation with various interventional radiology procedures[32,33], including regional and systemic therapy with various cytotoxic, biological, or targeted agents[34-37]. Therefore, PCT-CNB analysis of liver metastases may provide comprehensive insights for therapeutic GEP-NET decision-making.

Clinically, CT-guided biopsies enable risk stratification, as G3 upgrades may warrant therapy intensification[38,39]. Our study had several limitations. First, the study was retrospective and included a small number of stable G1 patients (n = 5); therefore, no survival comparisons were made between stable G1 patients and increased G1 patients. Second, while we observed significant associations between grade escalation and survival outcomes in the univariate analysis, multivariable adjustments for potential confounders (e.g., age, prior therapies) were not performed due to the limited subgroup sample sizes. However, baseline characteristics (Table 1) were not significantly different between the groups (all P > 0.05), suggesting that the observed survival differences are more likely attributable to tumor heterogeneity than to confounding factors. Third, while CT provides excellent anatomical information and improved deep lesion visualization compared with US or MRI, repeated CT scans deliver more ionizing radiation. Furthermore, the cohort predominantly comprised pancreatic-origin NETs (81.5%) from a single center, limiting generalizability to nonpancreatic primaries or diverse populations. Finally, “beam hardening” from biopsy needle metal artifacts can visually obscure target lesions.

CONCLUSION

In this study, PCT-CNB was safely performed with sufficient tumor sampling for reliable Ki-67 indexing and grading classifications of liver metastases from GEP-NETs. Our data revealed grade differences between liver tumors and primary site/secondary liver metastases. Critically, when the grade increased from G2, the OS rate significantly decreased. Future studies should validate these findings in prospective cohorts, explore genomic and epigenetic drivers of Ki-67 heterogeneity through multi-omics approaches, and compare biopsy modalities (e.g., CT-guided vs US-guided) to optimize clinical protocols.