Introduction

Gadoxetic acid (Eovist/Primovist [Bayer HealthCare])–enhanced MRI is highly sensitive and specific in detecting colorectal cancer (CRC) liver metastases [1,2,3,4,5] and is the preferred imaging examination to identify CRC metastases, determine surgical eligibility, and facilitate operative planning [3, 4]. Favorable diagnostic performance has also been demonstrated for liver metastases from other primary malignancies, including pancreatic adenocarcinoma (PDAC) and neuroendocrine tumors (NET) [1, 6,7,8].

Gadoxetic acid is a “hepatocyte-specific” MRI contrast agent that is eliminated via the urinary and biliary systems in approximately equal parts. As a result, gadoxetic acid has properties similar to other extracellular contrast agents, followed by active intracellular hepatocyte uptake and biliary excretion that peaks about 20 min after injection [9]. Gadoxetic acid is transported into liver cells mostly due to expression of organic anion transporter protein 1B3 (OATP1B3) [1, 10]. The high sensitivity for metastases on hepatobiliary phase (HBP) imaging is attributed to the contrast between the hyperintense background liver and hypointense non-hepatocyte-containing metastases. Metastases have classically been described as non-enhancing on HBP [1, 10].

Some reports in the radiologic literature have recognized “paradoxical” or “atypical” HBP enhancement in hepatic metastases from breast and colorectal primaries [11, 12]. These studies have mostly described a “target” appearance on HBP images, with central enhancement attributed to contrast accumulation within fibrotic and necrotic tissue. A target pattern is also seen in fibrotic hepatic neoplasms such as intrahepatic cholangiocarcinoma, as fibrotic tissues accumulate extracellular contrast [13, 14]. However, HBP enhancement is also associated with OATP1B3 expression in CRC metastases, indicating that HBP enhancement in non-hepatocyte-containing lesions is not solely attributed to fibrosis [15].

A number of malignancies, including CRC and PDAC, demonstrate OATP1B3 expression [16, 17]. Since OATP1B3 expression has prognostic significance in some malignancies, HBP enhancement might serve as a useful biomarker [18]. Although PDAC expresses OATP1B3 and is commonly imaged with gadoxetic acid, there is an absence of literature documenting corresponding HBP enhancement. Various patterns of HBP enhancement have been sporadically reported in the literature, including “target” and “mixed hypointense,” but require further exploration [11, 13, 14, 19]. The purpose of this study is to assess for and characterize hepatobiliary phase enhancement in hepatic metastases on gadoxetic acid–enhanced MRI.

Materials and methods

Patient cohort and study design

This was a retrospective study performed at a large quaternary care academic institution, with Institutional Review Board (IRB) approval and requirement for informed consent waived by the IRB. All data were collected and analyzed in compliance with the Health Insurance Portability and Accountability Act (HIPAA). All aspects of the study were performed in accordance with the Declaration of Helsinki.

Studies were retrieved from a search of our Picture Archiving and Communications System (PACS) for gadoxetic acid–enhanced MRI examinations performed between July 2012 and November 2019. Suitable studies were considered those with liver metastases. Exclusion criteria included as follows: lack of pathologic confirmation of hepatic metastases, isolated peripheral lesions that might be peritoneal, variable flip angles on pre-contrast and HBP images, severe motion artifacts on key sequences, and heterogeneous HBP enhancement of background liver parenchyma. Based on these criteria, out of 119 consecutive studies reviewed, 39 studies were excluded: 12 without pathologic confirmation, 3 with only superficial metastatic deposits (possibly peritoneal), 8 with different flip angles on pre-contrast and HBP images, 14 with motion artifacts that precluded quantitative analysis, and 2 with heterogeneous background liver parenchyma due to hemosiderosis. The final cohort for data analysis included 80 MRI studies. Basic demographic and clinical data was collected from medical records, including age, sex, and the temporal relationship of the MRI to systemic therapy (pre- or post-chemotherapy).

Imaging technique

All patients underwent gadoxetic acid–enhanced MRI of the abdomen on a 1.5- or 3-T MRI scanner at our institution with a phased array multi-channel body coil. Dynamic imaging was performed with axial 3-dimensional spoiled gradient-echo acquisitions with fat saturation after injection of 0.1 ml/kg gadoxetic acid. The pulse sequence parameters varied slightly by MRI machine, but representative parameters for the most common pre-contrast and dynamic post-contrast images are as follows: 1.5 T, liver acquisition with volume acquisition (LAVA), flip angle 15, TR 2.35/TE 1.05, 0.73 NEX, FOV 400 × 400 mm2, matrix 160 × 160, slice thickness 5 mm with 2.5 mm spacing. Delayed axial LAVA acquisitions were performed at 8 min with the same parameters as pre-contrast images, and at 20 min with either identical parameters or minimally increased TR/TE +/− slightly larger matrix (representative parameters for most common 20-min HBP acquisition: 1.5 T, LAVA, flip angle 15, TR 3.11/TE 1.44, 0.73 NEX, FOV 400 × 400 mm2, matrix 256 × 256, slice thickness 5 mm with 2.5 mm spacing. Most importantly, any additional 20-min acquisitions acquired with a higher flip angle (most commonly 25) were not used in quantitative assessment. To our knowledge, pre- and post-contrast images are not differentially scaled. Although not directly quantitatively assessed or measured, a routine abdominal MRI protocol was also performed on each patient that generally included as follows: coronal and axial SSFSE (single-shot fast spin echo), diffusion-weighted images, ADC, and axial T1 in and out of phase.

Image analysis

Images were retrospectively reviewed by two radiologists (R.B. and V.B.) to identify and measure hepatic metastases. Region-of-interest (ROI) measurements were placed on the dominant hepatic lesion (one dominant hepatic lesion per MRI), paraspinal musculature, and background liver on pre-contrast and HBP images. ROI on hepatic metastatic lesions were drawn to be greater than 25 mm2 for all lesions, at the location of highest intensity within the lesion. ROI measurements were used to calculate lesion-to-muscle and lesion-to-background liver ratios on pre-contrast and 20-min HBP images. To determine if lesions were enhancing and account for slight differences in pulse sequence parameters on HBP images, we assessed HBP enhancement relative to muscle (increase in lesion-to-muscle ratio), modified from a ratio used by Haimerl et al to quantify hepatic gadoxetic acid uptake [20]. HBP enhancement relative to muscle was calculated as follows:

$${\displaystyle \begin{array}{c}\mathrm{HBP}\ \mathrm{enhancement}\\ {}\mathrm{relative}\ \mathrm{to}\ \mathrm{muscle}\ \left(\%\right)\end{array}}=\frac{20-\min \mathrm{lesion}:\mathrm{muscle}-\mathrm{Pre}-\mathrm{contrast}\ \mathrm{lesion}:\mathrm{muscle}}{\mathrm{Pre}-\mathrm{contrast}\ \mathrm{lesion}:\mathrm{muscle}}$$

HBP enhancement relative to muscle > 20% was considered true enhancement. Although no threshold is defined in the literature regarding gadoxetic acid uptake, a 20% increase in ROI measurement to define enhancement was validated in renal lesions where pseudoenhancement did not exceed 13% (in pathologically proven cysts) and true enhancement exceeded 23% (in pathologically proven renal cell carcinomas) [21]. Previous studies assessing gadoxetic acid enhancement in metastatic lesions have used the lesion-to-background liver ratio on HBP (hepatobiliary phase relative intensity ratio [HBP RIR]) or increase in lesion-to-background liver ratio from pre-contrast to HBP images (hepatobiliary phase relative enhancement ratio [HBP RER]). We also calculated HBP RIR and HBP RER for dominant metastatic lesions.

Metastases were qualitatively subcategorized by pattern: peripheral, central (target), and diffuse heterogeneous. Lesions measuring < 1 cm were not subcategorized. On each MRI scan, the number of lesions demonstrating each pattern of uptake was counted, up to 10 for each pattern and 5 for subcentimeter lesions. The size of the largest lesion within each pattern was recorded.

Statistical analysis

Metastatic lesions were grouped by primary malignancy (CRC, n = 30; PDAC, n = 20; NET, n = 10; other, n = 20). For the three most common malignancies (CRC, PDAC, NET), characteristics were compared using non-parametric statistical tests. The Kruskal-Wallis test was used to compare quantitative measures of enhancement, number of lesions per MRI, and size of dominant lesions between groups. If a result was statistically significant, multiple comparisons were performed using Dunn’s test to compare the individual groups, and p values were adjusted with the Benjamin-Hochberg method (padj). Fisher’s exact test of independence was used to compare the proportion of dominant metastatic lesions that demonstrated enhancement, and the proportion of metastatic lesions that demonstrated each pattern of gadoxetic acid uptake (peripheral, central, diffuse) between groups. When significant results were obtained, pairwise comparisons were performed, and adjusted p values (padj) computed. Lesions with peripheral, central, and diffuse heterogeneous enhancement patterns were compared with respect to size, number of lesions, and HBP enhancement using the Kruskal-Wallis test. p values of less than 0.05 were considered to indicate a significant difference. All statistical analyses were performed in R (version 3.6.3; R Foundation for Statistical Computing).

Results

Patient demographics and descriptive statistics are summarized in Table 1. Of 80 patients, 44 were male (55%) and 36 were female (45%), with an average age of 60 years (range 27–84 years). Primary malignancies included CRC (n = 30), PDAC (n = 20), and NET (n = 10). The remainder (n = 20) had one of a heterogeneous group of primary malignancies (gastric, n = 5; medullary thyroid, n = 4; renal cell (RCC), n = 2; non–small cell lung cancer (NSCLC), n = 2; esophageal, n = 2; transitional cell, melanoma, chordoma, gastrointestinal stromal tumor (GIST), and leiomyosarcoma, n = 1 each). In total, 585 metastatic lesions were identified in these patients, including 202 CRC, 118 PDAC, 127 NET, and 138 other metastases.

Table 1 Descriptive statistics, stratified by primary malignancy. Number of MRI scans, number of metastases, demographic data, and temporal relationship of imaging examination to systemic therapy are noted
Full size table

Hepatobiliary phase enhancement

Most (77%) dominant lesions > 1 cm demonstrated true HBP enhancement, including most of CRC (74%), PDAC (73%), NET (100%), and other (72%) dominant metastases. In the “other” category, HBP enhancement was identified in dominant hepatic metastases from gastric (n = 4/5), esophageal (n = 2/2), medullary thyroid (n = 3/4), bladder (n = 1/1), RCC (n = 1/1), GIST (n = 1/1), and leiomyosarcoma (n = 1/1). No HBP enhancement was identified in cases of NSCLC (n = 2), melanoma (n = 1), or chordoma (n = 1).

For CRC, PDAC, and NET, quantitative measures of HBP enhancement, enhancement patterns, number of metastatic lesions per study, and size of dominant lesions are displayed in Table 2. Median HBP enhancement relative to muscle was lower for PDAC metastases (26%) compared with those for CRC (44%, padj = 0.04) and NET (51%, padj = 0.01) metastases. No statistically significant difference was detected between CRC and NET metastases (44 vs 51%, padj = 0.26). The distribution of HBP enhancement relative to muscle for metastases from each primary malignancy is displayed in Fig. 1.

Table 2 Characteristics of metastatic lesions, stratified by primary malignancy (colorectal cancer, pancreatic ductal adenocarcinoma, neuroendocrine tumor). Specifically, quantitative measures of enhancement in dominant lesions, pattern of enhancement for both dominant and all metastatic lesions, number of metastatic lesions per MRI, and size of dominant lesion are displayed for each group
Full size table
Fig. 1
figure 1

Boxplot of hepatobiliary phase (HBP) enhancement relative to muscle for liver metastases by primary malignancy. PDAC metastases had lower levels of enhancement on HBP phase than CRC and NET

Full size image

Median hepatobiliary phase relative intensity ratios (HBP RIR) for enhancing (0.47) and non-enhancing (0.53) CRC metastases were not significantly different (p = 0.96). PDAC metastases demonstrated a higher HBP RIR for enhancing lesions (0.48) compared with non-enhancing lesions (0.35) (p = 0.02). For both CRC and PDAC, hepatobiliary phase relative enhancement ratio (HBP RER) was significantly higher in enhancing lesions (75, 77%) than in non-enhancing lesions (66, 62%) (p = 0.01, 0.02).

Patterns of HBP enhancement

Patterns of enhancement observed, including peripheral, central/target, and diffuse heterogeneous patterns, are illustrated in Fig. 2. Primary analysis of dominant lesions demonstrated that proportions of each pattern of enhancement differed between CRC, PDAC, and NET (p < 0.001). In secondary analysis, PDAC metastases had a significantly higher proportion of the peripheral pattern (73%) than CRC (41%) and NET (8.9%) metastases (padj < 0.001 for both); NET metastases had a higher proportion of the central pattern (89%) compared with each of the other groups (CRC, 27%; PDAC, 18%; padj < 0.001 for both); and CRC metastases had a higher proportion of the diffuse heterogeneous pattern (32%) compared with PDAC (9.1%, padj = 0.006) and NET (2.5%, padj < 0.001) metastases, despite the peripheral pattern being the most common pattern identified in CRC metastases (41%).

Fig. 2
figure 2

Patterns of hepatobiliary phase (HBP) enhancement observed in metastases. Axial pre-gadolinium T1W image (a) and corresponding HBP image (b) in a 55-year-old man with neuroendocrine metastases demonstrate a central pattern of HBP enhancement (arrows). Axial pre-gadolinium T1W image (c) and coronal HBP image (d) in a 74-year-old man with pancreatic cancer demonstrate a peripheral pattern of HBP enhancement (arrows). Axial pre-gadolinium T1W image (e) and corresponding axial HBP image (f) in a 48-year-old woman with colorectal cancer demonstrate a diffuse pattern of enhancement. At quantitative assessment, the dominant lesion in all three of these patients demonstrated HBP enhancement

Full size image

Median dominant lesion size did not significantly differ between patterns of enhancement (peripheral = 28 mm; central = 18 mm; diffuse = 20 mm; p = 0.10). There was no difference in the median number of hepatic metastases per patient between patterns (7 peripheral; 8 central; 8 diffuse; p = 0.40). HBP enhancement relative to muscle was not equivalent between patterns of enhancement (p = 0.04). However, pairwise comparisons only demonstrated a tendency for the central pattern to enhance more than the diffuse (50 vs 26%, padj = 0.06) and peripheral patterns (50 vs 37%, padj = 0.12) that did not reach statistical significance.

Discussion

Hepatic metastases have classically been described as non-enhancing on hepatobiliary phase. But some studies have demonstrated HBP enhancement in metastases from breast and colon primaries, relating to fibrotic components in “target” lesions and correlating with OATP1B3 expression in CRC metastases [11, 12, 15]. OATP1B3 has potential prognostic significance, so HBP enhancement might serve as a biomarker in malignancies that express OATP1B3, including CRC and PDAC. But, to our knowledge, the presence of HBP enhancement has not been reported in most malignancies, including PDAC. We found HBP enhancement in hepatic metastases from 10 different primary malignancies, of which 77% of dominant lesions demonstrated HBP enhancement. For PDAC, 73% of dominant metastases demonstrated HBP enhancement, with levels of enhancement lower than CRC and NET metastases.

Patterns of HBP enhancement, illustrated in Fig. 2, varied by primary malignancy. PDAC and CRC metastases had the highest proportion of peripheral and diffuse HBP enhancement patterns, respectively, and NET metastatic lesions had the highest proportion of central (target) HBP enhancement. Both primary and secondary hepatic malignancies have a central (target) HBP enhancement pattern due to extracellular accumulation of contrast in areas of fibrosis [11, 13, 14, 19]. The pathologic correlate for non-central HBP enhancement in non-hepatocyte-containing lesions is less clear. Park et al demonstrated that HBP enhancement in “mixed hypointense” appearing CRC metastases correlated with OATP1B3 expression and had prognostic significance [15]. The predominance of non-central patterns of HBP enhancement in CRC and PDAC, both of which express OATP1B3, raises the possibility that non-central HBP enhancement might be due to OATP1B3 expression. However, this requires histologic correlation.

Despite differences in measurement technique, our finding that HBP enhancement is common in colorectal metastases is similar to those in other studies in the literature [12, 15, 19]. Most reports of HBP enhancement in metastatic lesions define enhancement in relation to background liver parenchyma [11, 15, 19]. Since liver parenchyma demonstrates variable uptake on HBP images [20], we felt this technique would be less reliable in assessing for true HBP enhancement. Our finding of HBP enhancement in hepatic metastases from PDAC, NET, and several other malignancies, to our knowledge, has not been reported. Radiologists should be aware of this phenomenon in hepatic metastases from primary tumors other than CRC and breast cancer.

The relative intensity of lesion-to-liver ratio on HBP (HBP RIR) likely correlates with a radiologist’s ability to perceive enhancement within a lesion, given that it measures contrast between the lesion and background liver. However, HBP RIR only correlated with true enhancement related to muscle in PDAC, and not in CRC. HBP RER (increase in HBP RIR from pre-contrast to HBP) correlated with true enhancement for both PDAC and CRC, suggesting that it is more reliable. Measurements of HBP RIR to assess for HBP enhancement should be avoided in future studies. Our findings suggest that our ability to perceive true enhancement on HBP images alone is limited, as we might expect given variable background liver uptake. All enhancing metastases in our series had a HBP RIR lower than 1, indicating they remained hypointense relative to background liver.

For CRC, studies have documented that HBP enhancement correlates with OATP1B3 expression, can help predict early treatment response [22], and is associated with improved overall survival [23]. Our novel finding of HBP enhancement in hepatic metastases from several other primary malignancies needs to be further explored, especially in PDAC which is known to express OATP1B3. Organic anion transporting polypeptides (OATP) are a family of transport proteins that have a role in the influx of a number of substrates into cells, including various pharmacologic agents used for chemotherapeutic treatment [24]. Both OATP1B1 and 1B3, for example, have a role in cellular uptake of irinotecan-based chemotherapy commonly used in CRC and PDAC [25]. The expression of transporters such as OATP1B3 has potential for use as a biomarker, as a drug target, and for prognostication [17, 25, 26]. Dedicated exploration of HBP enhancement in hepatic metastases from malignancies other than CRC is required, specifically assessing its relationship with pathologic findings, OAT1B3 expression levels, and prognosis. In PDAC, OATP1B3 expression is higher in primary lesions of earlier stage, so quantification of gadoxetic acid uptake in primary pancreatic lesions is also of interest [27].

There are a few limitations to our study. Histopathologic and clinical correlation was unavailable, and HBP enhancement could not be correlated with OATP1B3 expression and specific pathologic findings. This was a retrospective study performed at a single institution, which limits its generalizability. Although primary analyses were performed on single dominant lesions only, secondary analyses included several lesions per patient, which introduced clustering bias.

In conclusion, HBP enhancement was common in hepatic metastases from several primary malignancies and was identified in several malignancies where HBP enhancement has not been previously reported. Discrete patterns of HBP enhancement included peripheral, central (target), and diffuse heterogeneous patterns. The central (target) pattern, seen in primary and secondary tumors due to extracellular accumulation of contrast in fibrotic tissues, was more common in NET than in CRC and PDAC metastases. Non-central patterns were more common in CRC and PDAC, which are both known to express OATP1B3. Future studies are required to elucidate the relationship between HBP enhancement, enhancement pattern, OATP1B3 expression, and prognosis for PDAC and other primary malignancies.