Abstract
Purpose
To evaluate the diagnostic performance of 68Ga-DOTATATE PET, contrast-enhanced CT, combined DOTATATE PET/CT, and MRI in detecting neuroendocrine liver metastases on a per-lesion basis and to assess the added value of diagnostic contrast-enhanced CT to PET interpretation as well as influence of lesion size.
Materials and methods
This retrospective study evaluated patients with histologically-confirmed gastroenteropancreatic neuroendocrine tumors who underwent both contrast-enhanced MRI and 68Ga-DOTATATE PET/CT within 12 weeks between August 2017 and December 2023. Three readers evaluated in consensus MRI, 68Ga-DOTATATE PET, contrast-enhanced CT, and combined PET/CT in separate sessions. Lesions were stratified by size. Diagnostic performance metrics were calculated using generalized estimating equations with reference standard of imaging follow-up or histopathology.
Results
The study included 36 patients (mean age 66.4 ± 10.7 years, 55.6% male) with 720 lesions, of which 582 were metastases. MRI demonstrated superior performance (sensitivity 94.0%, specificity 94.2%) compared to all other modalities (all p < .001). Combined PET/CT showed superior sensitivity (68.1%) compared to PET alone (61.1%) (p < .001). The addition of diagnostic contrast-enhanced CT to PET interpretation identified 41 additional metastases. Size-stratified analysis revealed superior detection of subcentimeter lesions by MRI (sensitivity 83.3% for ≤ 5 mm, 99.6% for 6–10 mm) compared to PET/CT (34.8% and 78.6%, respectively; p < .001). For lesions > 10 mm, both MRI and PET/CT achieved 100% sensitivity, while PET alone reached 94.1% (p = .019).
Conclusion
MRI outperforms 68Ga-DOTATATE PET/CT in detecting individual neuroendocrine liver metastases, particularly for subcentimeter lesions. While diagnostic CT improves PET performance, the combination of MRI with 68Ga-DOTATATE PET/CT provides the most comprehensive assessment.
Keywords: Neuroendocrine tumor, Liver metastases, DOTATATE PET/CT, Magnetic resonance imaging, Diagnostic performance
Introduction
Neuroendocrine liver metastases (NELMs) develop in up to 90% of patients with gastroenteropancreatic neuroendocrine tumors (GEP-NETs), with hepatic tumor burden directly impacting survival rates and treatment options [1-3]. As the extent and number of metastases significantly influence prognosis and guide therapeutic decisions, accurate imaging assessment is crucial: five-year survival rates drop from 75% in patients with fewer than 10 metastases to 36% in those with extensive disease, and from 67 to 8% when liver involvement exceeds 50% of volume. Moreover, precise lesion detection directly impacts treatment planning, particularly in determining surgical candidacy and approach [4-8].
For patients with NELM, surgical management offers the best outcomes, with complete resection providing 5-year survival rates exceeding 60% [9]. Accurate preoperative imaging is essential for surgical planning, as candidates must be carefully selected based on tumor grade, metastatic distribution, and anticipated residual liver volume [10]. The importance of precise imaging extends beyond curativeintent surgery—even in palliative cases, accurate assessment of tumor burden guides surgical decision-making, as debulking procedures can significantly improve patient outcomes when substantial tumor reduction is achievable [11, 12].
There is no common agreement on optimal diagnostic imaging protocols for assessing liver metastases. Current imaging guidelines recommend contrast-enhanced computed tomography (CT), magnetic resonance imaging (MRI), and somatostatin receptor (SSTR) imaging for evaluating NELMs [13]. While previous research has examined different imaging approaches, studies comparing SSTR-PET/CT and MRI for NELM detection have shown mixed results in patient-based analyses. While each modality offers distinct advantages [14-17], their relative diagnostic performance for detecting individual metastases remains unclear, particularly regarding lesion detection across different size ranges. Previous studies comparing SSTR-PET/CT with MRI and CT have shown conflicting results on a per-lesion basis [18, 19]. Furthermore, the added value of diagnostic contrast-enhanced CT to DOTATATE PET interpretation has not been systematically evaluated on a per-lesion basis.
Therefore, this study aimed to compare the diagnostic performance of 68Ga-DOTATATE PET, contrast-enhanced CT, combined PET/CT, and MRI in detecting neuroendocrine liver metastases on a per-lesion basis. The primary objective was to evaluate the relative performance of each modality and assess how contrast-enhanced CT contributes to PET performance. Secondary objectives included analyzing the influence of size on detection across modalities, morphologic features on MRI and CT and functional parameters on PET imaging (standardized uptake values).
Methods
Study design and patient selection
A retrospective analysis was performed of patients with histologically confirmed gastroenteropancreatic neuroendocrine tumors (GEP-NETs) who underwent both liver MRI and contrast-enhanced 68Ga-DOTATATE PET/CT at a single institution between August 2017 and December 2023. The study was approved by the institutional review board (IRB-No: 2021P000432), and the requirement for informed consent was waived. A total of 355 consecutive patients were identified through the institutional radiology database. After initial screening, 140 patients were excluded due to no corresponding liver MRI within 12 weeks of 68Ga-DOTATATE PET. From the remaining 215 patients, further exclusions comprised: absence of liver metastases (n = 72), lack of follow-up or reference standard (n = 43), presence of another concurrent malignancy (n = 23), neuroendocrine carcinoma (n = 2), initiation of therapy during the imaging interval (n = 15), non-diagnostic image quality (n = 6), and absence of contrast-enhanced PET/CT (n = 18). The final study population consisted of 36 patients. The patient accrual flow chart is depicted in Fig. 1.
Fig. 1.
Patient accrual flowchart. GEP-NET = gastroenteropan-creatic neuroendocrine tumor; 68 Ga = gallium-68; PET = positron emission tomography; MRI = magnetic resonance imaging
MRI protocol
MRI examinations were conducted on 3 T systems (Discovery MR750, GE Healthcare, Milwaukee, WI; Magnetom Prisma Fit, Siemens Healthcare, Erlangen, Germany) using body array coils. The standardized liver protocol included volumetric T1-weighted Dixon imaging generating in-phase, opposed-phase, fat, and water images. Subsequently, respiratory-triggered T2-weighted sequences and diffusion-weighted imaging were performed with b-values of 0–50 and ≥ 800 s/mm2. Dynamic contrast enhancement utilized either gadoterate meglumine (Dotarem®, Guerbet; 0.1 mmol/kg) or gadoxetic acid (Eovist®/Primovist®, Bayer; 0.025 mmol/kg), administered via power injector at 1–2 mL/s followed by saline flush. The temporal acquisition scheme included pre-contrast imaging, late arterial phase (35–40 s), portal venous phase (60–75 s), and transitional phase (3–5 min) imaging. For gadoxetic acid studies, additional hepatobiliary phase images were acquired at 20 min post-contrast administration. Detailed technical parameters are listed in Table S1 of the supplemental material.
PET/CT protocol
Integrated PET/CT examinations were performed on time-of-flight-enabled systems (Siemens Vision 600, Siemens mCT, Knoxville, TN). Patients received intravenous 68Ga-DOTATATE with weight-adjusted dosing (0.054 mCi/kg, not exceeding 5.4 mCi). PET acquisition commenced 60 min post-radiotracer administration, extending from vertex to mid-thighs. The CT protocol incorporated initial non-contrast imaging followed by weight-based administration of iodinated contrast (1.5 ml Iomeprol 400 mgI/mL; Iomeron®, Bracco) via power injector. Post-contrast imaging included arterial phase (35–40 s) and portal venous phase (60–75 s) acquisitions, with contrast delivery followed by a 30 mL saline flush. Detailed technical parameters are listed in Table S2 of the supplemental material.
Image analysis
All imaging studies underwent consensus review by a team of three readers at different training levels: a senior staff radiologist combining extensive experience in both abdominal and nuclear medicine imaging (25 and 14 years respectively), a diagnostic imaging fellow with nuclear medicine training (6 years), and a radiology resident (5 years). Image interpretation was performed using specialized medical imaging software (Visage Client Version 7.1.18, Visage Imaging, Inc., San Diego, CA) on high-resolution, DICOM-calibrated diagnostic workstations. To minimize bias, readers had access only to the confirmed NET diagnosis, while all other clinical information was withheld during the evaluation process.
A systematic three-phase evaluation process was implemented to ensure comprehensive and unbiased assessment. In the initial phase, separate reading sessions for PET, CT, PET/CT, and MRI datasets were conducted in randomized order. A minimum four-week interval between sessions was maintained to minimize recall bias. Lesion analysis included measurement of maximum axial diameter (mm) and assessment of modality-specific imaging features.
Enhancement patterns on CT and MRI were characterized by the presence or absence of arterial phase hyperenhancement and portal venous phase washout. For PET analysis, lesions were considered positive when demonstrating focal DOTATATE uptake visually exceeding liver background. Quantitative PET parameters included lesion maximum standardized uptake values (SUVmax), normal right liver parenchyma mean standardized uptake values (SUVmean) sampled from regions without focal uptake abnormalities, and mediastinal blood pool SUVmean measured at the ascending aorta. These measurements enabled calculation of tumor-to-liver (TLR) and tumor-to-blood pool ratios (TBR).
The second phase established lesion correspondence across all imaging modalities through consensus review. In cases where unanimous agreement could not be reached through joint review and discussion among the three readers, the senior reader’s assessment was considered definitive. The final phase compared identified lesions against the reference standard of histopathologic confirmation or imaging follow-up when tissue sampling was not feasible.
Statistical analysis
Statistical analysis was conducted using SPSS Statistics version 29 (IBM Corp., Armonk, NY, USA) and CIA (Confidence Interval Analysis©, Trevor Bryant, University of Southampton). Data distribution was assessed using Shapiro–Wilk test. Descriptive statistics presented continuous variables as mean ± standard deviation when normally distributed and median (interquartile range) when non-normally distributed. Categorical variables were reported as absolute numbers and percentages.
Per-lesion diagnostic performance metrics including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated with corresponding 95% confidence intervals. Given the clustering of multiple lesions within individual patients, Generalized Estimating Equations were employed to compare diagnostic performance between imaging modalities (PET, CT, combined PET/CT, MRI without hepatobiliary phase, and MRI including hepatobiliary phase).
To investigate size-dependent detection patterns, lesions were stratified into subgroups of ≤ 5 mm, 6-10 mm and ≥ 10 mm diameter. Intermodality comparisons of detection rates within size subgroups were performed using Generalized Estimating Equations with Bonferroni correction for multiple comparisons. Statistical significance was set at p < 0.05.
Results
Patient and lesion characteristics
A total of 36 patients (mean age 66.4 ± 10.7 years, 55.6% male) were included in the analysis. Primary tumor sites were predominantly pancreatic (52.8%) and small bowel/stomach (47.2%). Tumor grading revealed NET G1 in 41.7%, NET G2 in 55.6%, and NET G3 in 2.8% of cases. Prior systemic treatment had been administered in 24.3% of patients, including somatostatin analogues alone (16.7%) and somatostatin analogues combined with chemotherapy, peptide receptor radionuclide therapy, transarterial embolization, or tyrosine kinase inhibitors (8.3%). Additionally, 13.9% of patients had undergone prior partial hepatectomy or local ablative therapy. A total of 720 lesions were analyzed, of which 582 were confirmed metastases, either on follow-up imaging (n = 534) or histopathology (n = 48). Among the metastases, the majority were NET G2 (69.9%), followed by NET G1 (26.1%) and NET G3 (4.0%). The median number of metastases per patient was 11.5 (IQR: 2.0–17.2) with a median lesion size of 6 mm (IQR: 5–10 mm). Table 1 includes detailed patient and lesion characteristics.
Table 1.
Patient and lesion characteristics
| Characteristic | Patients (N = 36) | Metastases (N = 582) |
|---|---|---|
| Age | 66.4 ± 10.7 years | – |
| Gender (Male) | 20 (55.6%) | – |
| Primary site | ||
| Pancreatic | 19 (52.8%) | 378 (65%) |
| Small Bowel/Stomach | 17 (47.2%) | 204 (35%) |
| Tumor grading | ||
| NET G1 | 15 (41.7%) | 152 (26.1%) |
| NET G2 | 20 (55.6%) | 407 (69.9%) |
| NET G3 | 1 (2.8%) | 23 (4.0%) |
| Prior systemic treatment | 9 (25%) | 269 (46.2%) |
| SSA | 6 (66.7%) | 216 (80.3%) |
| SSA + Chemotherapy, PRRT, TAE or TKI | 3 (33.3%) | 53 (19.7%) |
| Prior partial hepatectomy or local ablative therapy | 5 (13.9%) | – |
| Reason for imaging | ||
| Staging | 16 (44.4%) | 186 (32%) |
| Restaging | 20 (55.6%) | 396 (68%) |
| Lesion characteristics | ||
| Total metastases | 582 (81.8%) | – |
| Metastases per patient | 11.5 (2.0–17.2) | – |
| Total benign lesions | 138 (19.2%) | – |
| Benign lesions per patient | 4 (3.0–5.0) | – |
| Lesion size | – | 6.0 mm (5.0–10) |
| PET parameters | ||
| SUVmax | – | 13.0 (9.7–19.4) |
| Tumor-to-liver ratio | – | 2.0 (1.5–3.3) |
| Tumor-to-blood pool ratio | – | 13.7 (10.1–21.2) |
| Liver background SUVmean | 6.5 (4.8–8.5) | – |
| Blood pool SUVmean | 0.9 (0.3) | – |
| Contrast media | ||
| Gadoterate meglumine | 31 (86.1%) | 487 (83.7%) |
| Gd-EOBDTPA | 5 (13.9%) | 95 (16.3%) |
Patient and imaging characteristics for the study population (N = 36) and analyzed metastases (N = 582). Continuous variables are presented as mean ± standard deviation for normally distributed data (age) or median (interquartile range) for non-normally distributed data. Categorical variables are presented as numbers (percentages). NET = neuroendocrine tumor; SSA = somatostatin analogues; PRRT = peptide receptor radionuclide therapy; TAE = transarterial embolization; TKI = tyrosine kinase inhibitor; PET = positron emission tomography; SUV = standardized uptake value; Gd-EOBDTPA = gadoxetic acid
Imaging characteristics
PET-positive metastases were significantly larger than PET-negative lesions (median: 8 mm [IQR: 6.0–12.0] vs. 5.0 mm [IQR: 4.0–6.0], p < 0.0001). Similar size differences were observed for PET/CT (positive: 8.0 mm [IQR: 6.0–12.0] vs. negative: 5.0 mm [IQR: 4.0–6.0], p < 0.0001) and CT (positive: 10.0 mm [IQR: 7.0–15.8] vs. negative: 6.0 mm [IQR: 4.0–7.0], p < 0.0001). MRI showed the smallest size difference between detected and undetected lesions (positive: 7.0 mm [IQR: 5.0–10.0] vs. negative: 4.0 mm [IQR: 3.2–4.0], p < 0.0001). Arterial phase detection rate was significantly higher on MRI compared to CT (68.2 vs. 31.2%, p < 0.001), as was arterial enhancement (61.2 vs. 28.3%, p < 0.001). Portal venous phase detection rate was also higher on MRI (29.6 vs. 22.7%, p < 0.001), while portal venous washout was similar between modalities (12.5 vs. 12.4%, p = 1.000). For PET-positive metastases, the median SUVmax was 13.0 (IQR: 9.7–19.4), median TLR was 2.0 (IQR: 1.5–3.3), and median TBR was 13.7 (IQR: 10.1–21.2). PET parameters demonstrated size-dependent differences (all p < 0.001). Lesions ≤ 5 mm showed lower SUVmax values (median: 10.2 [IQR: 8.9–12.0]) compared to both 6–10 mm lesions (median: 11.9 [IQR: 9.0–15.5], p = 0.0058) and > 10 mm lesions (median: 22.1 [IQR: 14.7–34.0], p < 0.001). The difference in SUVmax between 6-10 mm and > 10 mm lesions was also notable (p < 0.001). Similar patterns were observed for TLR, with values increasing from ≤ 5 mm lesions (median: 1.5 [IQR: 1.4–1.8]) to 6–10 mm lesions (median: 1.8 [IQR: 1.5–2.5], p = 0.0001) and > 10 mm lesions (median: 3.7 [IQR: 2.5–6.0], p < 0.001). For TBR, while there was no difference between ≤ 5 mm (median: 12.1 [IQR: 9.6–14.1]) and 6–10 mm lesions (median: 12.3 [IQR: 9.0–16.1], p = 0.3887), both groups showed lower values compared to lesions > 10 mm (median: 22.9 [IQR: 15.1–36.7], both p < 0.001).
Diagnostic performance
MRI including HBP demonstrated a sensitivity of 94.0% (95% CI 91.8–95.6%), specificity of 94.2% (95% CI 89.0–97.0%), PPV of 98.6% (95% CI 97.2–99.3%), NPV of 78.8% (95% CI 71.9–84.3%), and diagnostic accuracy of 94.0% (95% CI 92.1–95.5%). MRI without HBP showed similar performance with sensitivity of 93.7% (95% CI 91.4–95.4%), specificity of 93.5% (95% CI 88.1–96.5%), PPV of 98.4% (95% CI 96.9–99.1%), NPV of 77.7% (95% CI 70.8–83.4%), and accuracy of 93.6% (95% CI 91.6–95.2%).
PET/CT demonstrated a sensitivity of 68.1% (95% CI 64.1–71.7%), specificity of 81.8% (95% CI 74.4–87.4%), PPV of 94.1% (95% CI 91.4–95.9%), NPV of 37.6% (95% CI 32.3–43.2%), and accuracy of 70.7% (95% CI 67.3–73.9%). Compared to PET alone, which showed a sensitivity of 61.1% (95% CI 57.0–64.9%), specificity of 81.2% (95% CI 73.8–86.8%), PPV of 93.2% (95% CI 90.2–95.3%), NPV of 33.0% (95% CI 28.2–38.2%), and accuracy of 64.9% (95% CI 61.3–68.3%), PET/CT demonstrated significantly better sensitivity, accuracy, PPV, and NPV (all p < 0.001), with no significant difference in specificity (p = 0.421). MRI without HBP significantly outperformed PET/CT across all diagnostic metrics (all p < 0.001). CT demonstrated a sensitivity of 35.8% (95% CI 32.0–39.8%), specificity of 68.1% (95% CI 59.9–75.3%), PPV of 82.6% (95% CI 77.2–87.0%), NPV of 20.1% (95% CI 16.6–24.1%), and accuracy of 42.0% (95% CI 38.4–45.7%). MRI without HBP, PET/CT, and PET alone all significantly outperformed CT across all diagnostic metrics (all p < 0.001) (Table 2).
Table 2.
Diagnostic performance of imaging modalities
| Modality | Sensitivity | Specificity | PPV | NPV | Accuracy |
|---|---|---|---|---|---|
| CT | 35.8% (32.0–39.8%) | 68.1% (59.9–75.3%) | 82.6% (77.2–87.0%) | 20.1% (16.6–24.1%) | 42.0% (38.4–45.7%) |
| PET | 61.1% (57.0–64.9%) | 81.2% (73.8–86.8%) | 93.2% (90.2–95.3%) | 33.0% (28.2–38.2%) | 64.9% (61.3–68.3%) |
| PET/CT | 68.1% (64.1–71.7%) | 81.8% (74.4–87.4%) | 94.1% (91.4–95.9%) | 37.6% (32.3–43.2%) | 70.7% (67.3–73.9%) |
| MRI without HBP | 93.7% (91.4–95.4%) | 93.5% (88.1–96.5%) | 98.4% (96.9–99.1%) | 77.7% (70.8–83.4%) | 93.6% (91.6–95.2%) |
| MRI including HBP | 94.0% (91.8–95.6%) | 94.2% (89.0–97.0%) | 98.6% (97.2–99.3%) | 78.8% (71.9–84.3%) | 94.0% (92.1–95.5%) |
Diagnostic performance metrics of different imaging modalities in detecting neuroendocrine liver metastases. Data are presented as percentages with 95% confidence intervals in parentheses. CT = computed tomography; PET = positron emission tomography; MRI = magnetic resonance imaging; HBP = hepatobiliary phase; PPV = positive predictive value; NPV = negative predictive value
Size-stratified sensitivity
Of 583 metastases, 198 (34.0%) were ≤ 5 mm, 262 (44.9%) were 6–10 mm, and 123 (21.1%) were > 10 mm. For lesions ≤ 5 mm, MRI including HBP and MRI without HBP showed sensitivities of 83.3% [95% CI 77.5–87.9%] and 82.3% [95% CI 76.4–87.0%], respectively, outperforming all other modalities (all p < 0.001). PET/CT (34.8% [95% CI 28.6–41.7%]) demonstrated superior sensitivity compared to PET alone (27.8% [95% CI 22.0–34.4%], p < 0.001) and CT (10.1% [95% CI 6.6–15.1%], p < 0.001). In the 6–10 mm category, both MRI protocols demonstrated a sensitivity of 99.6% [95% CI 97.9–99.9%], surpassing all other modalities (all p < 0.001). PET/CT detected 78.6% [95% CI 73.3–83.2%], exceeding PET alone (71.0% [95% CI 65.2–76.2%], p < 0.001) and CT (34.0% [95% CI 28.5–39.9%], p < 0.001). For lesions > 10 mm, both MRI protocols and PET/CT achieved 100% sensitivity [95% CI 96.8–100.0%], while PET alone showed slightly lower sensitivity (94.1% [95% CI 88.3–97.1%], p = 0.019) and CT demonstrated the lowest sensitivity (82.2% [95% CI 74.3–88.1%], p < 0.001 compared to all other modalities). Comparing size groups, sensitivity improved with increasing lesion size for all modalities (p < 0.001), except for both MRI protocols which showed no difference between lesions measuring 6–10 mm versus > 10 mm (p = 0.949). Detailed information about the size-stratified analysis for each modality are presented in Table 3 and Fig. 2.
Table 3.
Size-stratified sensitivity analysis for true metastases
| Modality | ≤ 5 mm (n = 198) | 6–10 mm (n = 262) | > 10 mm (n = 118) |
|---|---|---|---|
| CT | 10.1% (6.6–15.1%) | 34.0% (28.5–39.9%) | 82.2% (74.3–88.1%) |
| PET | 27.8% (22.0–34.4%) | 71.0% (65.2–76.2%) | 94.1% (88.3–97.1%) |
| PET/CT | 34.8% (28.6–41.7%) | 78.6% (73.3–83.2%) | 100.0% (96.8–100.0%) |
| MRI without HBP | 82.3% (76.4–87.0%) | 99.6% (97.9–99.9%) | 100.0% (96.8–100.0%) |
| MRI including HBP | 83.3% (77.5–87.9%) | 99.6% (97.9–99.9%) | 100.0% (96.8–100.0%) |
Size-stratified sensitivity analysis for the detection of neuroendocrine liver metastases across different imaging modalities. Data are presented as percentages with 95% confidence intervals in parentheses. Results demonstrate the impact of lesion size on detection sensitivity for each imaging modality. CT = computed tomography; PET = positron emission tomography; MRI = magnetic resonance imaging; HBP = hepatobiliary phase
Fig. 2.
Size-dependent detection rates of imaging modalities for neuroendocrine liver metastases. Sensitivity analysis stratified by lesion size demonstrates superior detection by MRI (with/without HBP) across all size categories compared to CT and PET/CT. Detection rates improved with increasing lesion size for all modalities, with particularly pronounced differences in lesions ≤ 5 mm. Error bars: 95% CI
Modality agreement
In the comparison between modalities, 41 metastases (7.0%) were identified by PET/CT but missed on PET alone. Between CT and MRI, 341 metastases (58.5%) were visualized by MRI but not by CT, while 2 lesions (0.3%) were seen on CT but not on MRI. The comparison between MRI and PET/CT revealed that 157 metastases (26.9%) were identified by MRI but missed on PET/CT, while 6 lesions (1.0%) were PET/CT-positive but MRI-negative (Figs. 3 and 4).
Fig. 3.
59-year-old man with pancreatic neuroendocrine tumor (G2). a Diffusion-weighted imaging (b = 800 s/mm2), b arterial phase MRI, c 68Ga-DOTATATE PET, and d arterial phase CT images show multiple hepatic metastases. Thin arrows indicate metastases in segment VIII visible on all imaging modalities. Arrowheads demonstrate metastases that appear less conspicuous on PET but are confidently characterized as metastases in combination with arterial enhancement on CT. Thick arrow indicates a small metastasis detected only on MRI, missed on both PET and CT
Fig. 4.
68-year-old woman with neuroendocrine tumor (G1) of the ileum. Multiple hepatic metastases demonstrated on a diffusion-weighted imaging (b = 800 s/mm2), b corresponding ADC map, c 68Ga-DOTATATE PET, and d arterial phase CT. Thin arrow points to metastases in liver segment V visible on all modalities. Arrowhead indicates a tiny metastasis showing arterial enhancement on CT and restricted diffusion on DWI/ADC; while this lesion demonstrates limited DOTATATE uptake on PET alone, it is confidently classified as metastasis on combined PET/CT interpretation
Discussion
In this retrospective study, we performed a comprehensive per-lesion analysis comparing the diagnostic performance of current imaging modalities for neuroendocrine liver metastases. To our knowledge, this study represents the largest cohort comparing contrast-enhanced DOTATATE PET/ CT with MRI on a per-lesion basis, and is the first to systematically evaluate the added value of diagnostic contrast-enhanced CT to PET interpretation, demonstrating MRI’s superior sensitivity of 94.0% particularly for subcentimeter lesions, while contrast-enhanced CT improved PET performance compared to PET alone.
Prior studies evaluating the comparative performance of SSTR-PET/CT and conventional imaging for NELM detection have shown conflicting results. One recent study analyzing 32 patients with 90 liver metastases found that both DWI and DCE-MRI detected significantly more lesions than DOTATATE PET/CT, reporting lower sensitivities for subcentimeter lesions on PET (42%) compared to MRI sequences (DWI 67%, DCE 76%) [18]. These findings align with our results, though our larger cohort allowed for a more detailed size-stratified analysis and comprehensive evaluation of enhancement patterns and PET parameters across different lesion sizes.
Contrasting results were reported in a different analysis of 32 patients that suggested superior performance of DOTATATE PET compared to conventional imaging [19]. However, their methodology differed substantially, combining CT and MRI findings with disproportionate use of CT (26 versus 17 patients), and lacking standardized MRI protocols. A larger prospective evaluation of 109 patients comparing DOTANOC PET/CT to conventional imaging reported high sensitivity (97.4%) for metastatic disease, with PET detecting more metastatic sites overall compared to conventional imaging (106 vs. 77 regions) [20]. However, their analysis did not specify the distribution of imaging modalities used nor provide detailed analysis specific to liver metastases.
Different methodological approaches have provided additional insights into the relative strengths of these imaging modalities. One study focused on comparing dynamic contrast-enhanced MRI perfusion parameters with PET parameters in 32 patients with hepatic metastases from neuroendocrine neoplasms [21]. Their analysis demonstrated high diagnostic accuracy for both PET parameters (SUVmean) and DCE-MRI perfusion parameters in distinguishing metastases from liver tissue, suggesting complementary roles of functional parameters from both modalities. Another study evaluated integrated PET/MRI, demonstrating sensitivity of 89.8% compared to 81.6% for PET/CT in detecting metastatic disease, though their protocol did not include diffusion-weighted imaging sequences [22].
The findings of this study demonstrate the complementary role of SSTR-PET/CT and MRI in the assessment of NELMs, while highlighting the superior detection capability of MRI, particularly for small lesions. The significantly lower sensitivity of PET and PET/CT for subcentimeter lesions emphasizes the importance of including MRI in the initial staging workup, as missing small metastases could impact treatment planning, particularly in patients being considered for surgical resection. While the addition of diagnostic contrast-enhanced CT improved PET performance by detecting 41 additional metastases, this improvement was limited compared to the detection capability of MRI. Furthermore, the strong size dependence of PET parameters suggests that SSTR expression quantification may be less reliable in smaller lesions. Notably, approximately 7% of metastases in our cohort were hypovascular based on the difference between arterial phase detection and enhancement rates. For hypovascular lesions, multiparametric MRI including diffusion-weighted imaging and hepatobiliary phase imaging, as well as DOTATATE PET, may provide added diagnostic value by offering complementary information independent of vascular enhancement patterns, thereby improving detection of atypical neuroendocrine liver metastases. Our results support a multimodality imaging approach, with MRI being essential for accurate lesion detection and PET/CT providing additional functional information through SSTR expression assessment, which remains valuable for treatment planning, particularly regarding the potential role of peptide receptor radionuclide therapy.
Several limitations of this retrospective analysis should be acknowledged. Although data were collected consecutively to minimize selection bias, and our cohort represents one of the largest evaluations of NELMs detection across multiple imaging modalities on a per-lesion basis, multicenter validation would strengthen the generalizability of our findings.
The most fundamental limitation, inherent to imaging studies evaluating multifocal metastatic disease, was the inability to obtain histopathological confirmation for each lesion, as biopsy of all hepatic lesions would not be ethically or practically feasible. To address this limitation consensus reading was performed and follow-up imaging was used as the reference standard, in case histopathology was unavailable.
Another methodological limitation is the temporal separation between MRI and PET/CT examinations, although it is limited to a maximum of 12 weeks. However, due to the slow progression rate typical of neuroendocrine tumors and the fact that only 25% of our study population were receiving active systemic treatment at the time of imaging, there was likely no substantial difference during this time frame. Additionally, we excluded patients who were started on systemic therapy during the time interval between imaging.
Furthermore, our DWI protocol utilized a minimum b-value threshold of 800 s/mm2, which may be suboptimal by current 3 T standards, where b-values ≥ 1000 s/mm2 might be preferred. This reflects the evolution of imaging protocols during our study period from 2017 to 2023. However, the reduction in signal-to-noise ratio associated with higher b-values might also downplay lesion detectability [23].
Of note, our CT protocol was part of a PET/CT examination rather than a dedicated liver study. Currently, there are no standardized acquisition protocols for PET/CT, and several institutions do not even acquire diagnostic-quality CT; instead, they only obtain attenuation correction CT during their PET/CT studies. Our PET/CT protocol includes diagnostic-quality arterial and portal venous phase CT scans, which are not technically inferior to standalone CT in detecting small hypervascular lesions. Notably, optimized acquisition parameters, use of dual-energy CT, or newer photon-counting scanners could have potentially improved CT sensitivity, which should be considered when interpreting our results.
Lastly, although our study demonstrated superior MRI detection rates, particularly for small lesions, the clinical impact of identifying these additional metastases was not assessed. Such an analysis would require consideration of multiple factors beyond imaging findings, including clinical symptoms, overall disease burden, extrahepatic manifestations, and treatment history.
In conclusion, our study demonstrated the superiority of MRI compared to 68 Ga-DOTATATE PET/CT in the detection of individual neuroendocrine liver metastases. While the addition of diagnostic contrast-enhanced CT to PET improved sensitivity and specificity, the detectability for subcentimeter lesions remains moderate compared to MRI´s performance. This strengthens the recommendation of combining diagnostic PET/CT and liver MRI for neuroendocrine tumor staging.
Supplementary Material
Supplementary Information The online version contains supplementary material available at 10.1007/s11547-025-02080-1.
Funding
The authors have not disclosed any funding.
Abbreviations
- CI
Confidence interval
- CT
Computed tomography
- DCE
Dynamic contrast-enhanced
- DWI
Diffusion-weighted imaging
- 68Ga
Gallium-68
- GEP-NET
Gastroenteropancreatic neuroendocrine tumor
- HBP
Hepatobiliary phase
- IQR
Interquartile range
- MRI
Magnetic resonance imaging
- NELM
Neuroendocrine liver metastasis
- NET
Neuroendocrine tumor
- NPV
Negative predictive value
- PET
Positron emission tomography
- PPV
Positive predictive value
- SSTR
Somatostatin receptor
- SUVmax
Maximum standardized uptake value
- SUVmean
Mean standardized uptake value
- TBR
Tumor-to-blood pool ratio
- TLR
Tumor-to-liver ratio
Footnotes
Conflict of interest The authors have not disclosed any competing interests.
Ethical approval This retrospective study, involving the analysis of anonymized patient data, was conducted in accordance with the ethical standards of the institutional research ethical committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study protocol was reviewed and approved by the Institutional Review Board (IRB) of the Mass General Brigham. Given the retrospective nature of the study, the use of de-identified data, and the minimal risk to patient privacy, the requirement for individual informed consent was formally waived by the IRB. All data were handled with strict adherence to patient confidentiality and privacy regulations.
References
- 1.Pape UF, Berndt U, Müller-Nordhorn J et al. (2008) Prognostic factors of long-term outcome in gastroenteropancreatic neuroendocrine tumours. Endocr Relat Cancer 15(4):1083–1097. 10.1677/ERC-08-0017 [DOI] [PubMed] [Google Scholar]
- 2.Pavel M, Baudin E, Couvelard A et al. (2012) ENETS consensus guidelines for the management of patients with liver and other distant metastases from neuroendocrine neoplasms of foregut, midgut, hindgut, and unknown primary. Neuroendocrinology 95(2):157–176. 10.1159/000335597 [DOI] [PubMed] [Google Scholar]
- 3.Riihimäki M, Hemminki A, Sundquist K, Sundquist J, Hemminki K (2016) The epidemiology of metastases in neuroendocrine tumors. Int J Cancer 139(12):2679–2686. 10.1002/ijc.30400 [DOI] [PubMed] [Google Scholar]
- 4.Thompson GB, van Heerden JA, Grant CS, Carney JA, Ilstrup DM (1988) Islet cell carcinomas of the pancreas: a twenty-year experience. Surgery 104(6):1011–1017 [PubMed] [Google Scholar]
- 5.Godwin JD (1975) Carcinoid tumors. An analysis of 2,837 cases. Cancer 36(2):560–569 [DOI] [PubMed] [Google Scholar]
- 6.Mayo SC, de Jong MC, Pulitano C et al. (2010) Surgical management of hepatic neuroendocrine tumor metastasis: results from an international multi-institutional analysis. Ann Surg Oncol 17(12):3129–3136. 10.1245/s10434-010-1154-5 [DOI] [PubMed] [Google Scholar]
- 7.Touzios JG, Kiely JM, Pitt SC et al. (2005) Neuroendocrine hepatic metastases: does aggressive management improve survival? Ann Surg 241(5):776–785. 10.1097/01.sla.0000161981.58631.ab [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Durante C, Boukheris H, Dromain C et al. (2009) Prognostic factors influencing survival from metastatic (stage IV) gastroenteropancreatic well-differentiated endocrine carcinoma. Endocr Relat Cancer 16(2):585–597. 10.1677/ERC-08-0301 [DOI] [PubMed] [Google Scholar]
- 9.Fairweather M, Swanson R, Wang J et al. (2017) Management of neuroendocrine tumor liver metastases: long-term outcomes and prognostic factors from a large prospective database. Ann Surg Oncol 24(8):2319–2325. 10.1245/s10434-017-5839-x [DOI] [PubMed] [Google Scholar]
- 10.O’Toole D, Kianmanesh R, Caplin M (2016) ENETS 2016 consensus guidelines for the management of patients with digestive neuroendocrine tumors: an update. Neuroendocrinology 103(2):117–118. 10.1159/000443169 [DOI] [PubMed] [Google Scholar]
- 11.Mahuron KM, Singh G (2023) Defining a new classification system for the surgical management of neuroendocrine tumor liver metastases. J Clin Med 12(7):2456. 10.3390/jcm12072456 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Cloyd JM, Ejaz A, Konda B, Makary MS, Pawlik TM (2020) Neuroendocrine liver metastases: a contemporary review of treatment strategies. Hepatobiliary Surg Nutr 9(4):440–451. 10.21037/hbsn.2020.04.02 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Shah MH, Goldner WS, Benson AB et al. (2021) Neuroendocrine and adrenal tumors, version 2.2021, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw 19(7):839–868. 10.6004/jnccn.2021.0032 [DOI] [PubMed] [Google Scholar]
- 14.Dromain C, de Baere T, Lumbroso J et al. (2005) Detection of liver metastases from endocrine tumors: a prospective comparison of somatostatin receptor scintigraphy, computed tomography, and magnetic resonance imaging. J Clin Oncol 23(1):70–78. 10.1200/JCO.2005.01.013 [DOI] [PubMed] [Google Scholar]
- 15.d’Assignies G, Fina P, Bruno O et al. (2013) High sensitivity of diffusion-weighted MR imaging for the detection of liver metastases from neuroendocrine tumors: comparison with T2-weighted and dynamic gadolinium-enhanced MR imaging. Radiology 268(2):390–399. 10.1148/radiol.13121628 [DOI] [PubMed] [Google Scholar]
- 16.Lavelle LP, O’Neill AC, McMahon CJ et al. (2016) Is diffusion-weighted MRI sufficient for follow-up of neuroendocrine tumour liver metastases? Clin Radiol 71(9):863–868. 10.1016/j.crad.2016.05.016 [DOI] [PubMed] [Google Scholar]
- 17.Morin C, Drolet S, Daigle C et al. (2020) Additional value of gadoxetic acid-enhanced MRI to conventional extracellular gadolinium-enhanced MRI for the surgical management of colorectal and neuroendocrine liver metastases. HPB (Oxford) 22(5):710–715. 10.1016/j.hpb.2019.09.009 [DOI] [PubMed] [Google Scholar]
- 18.Haider M, Jiang BG, Parker JA, Bullock AJ, Goehler A, Tsai LL (2022) Use of MRI and Ga-68 DOTATATE for the detection of neuroendocrine liver metastases. Abdom Radiol (NY) 47(2):586–595. 10.1007/s00261-021-03341-z [DOI] [PubMed] [Google Scholar]
- 19.Jackson T, Darwish M, Cho E, Nagatomo K, Osman H, Jeyarajah DR (2021) 68Ga-DOTATATE PET/CT compared to standard imaging in metastatic neuroendocrine tumors: a more sensitive test to detect liver metastasis? Abdom Radiol (NY) 46(7):3179–3183. 10.1007/s00261-021-02990-4 [DOI] [PubMed] [Google Scholar]
- 20.Naswa N, Sharma P, Kumar A et al. (2011) Gallium-68-DOTANOC PET/CT of patients with gastroenteropancreatic neuroendocrine tumors: a prospective single-center study. AJR Am J Roentgenol 197(5):1221–1228. 10.2214/AJR.11.7298 [DOI] [PubMed] [Google Scholar]
- 21.Armbruster M, Zech CJ, Sourbron S et al. (2014) Diagnostic accuracy of dynamic gadoxetic-acid-enhanced MRI and PET/CT compared in patients with liver metastases from neuroendocrine neoplasms. J Magn Reson Imaging 40(2):457–466. 10.1002/jmri.24363 [DOI] [PubMed] [Google Scholar]
- 22.Berzaczy D, Giraudo C, Haug AR et al. (2017) Whole-body 68Ga-DOTANOC PET/MRI versus 68Ga-DOTANOC PET/CT in patients with neuroendocrine tumors: a prospective study in 28 patients. Clin Nucl Med 42(9):669–674. 10.1097/RLU.0000000000001753 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Goshima S, Kanematsu M, Kondo H et al. (2008) Diffusion-weighted imaging of the liver: optimizing b value for the detection and characterization of benign and malignant hepatic lesions. J Magn Reson Imaging 28(3):691–697. 10.1002/jmri.21467 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.