Acta Radiol OnlineFirst, published on November 29, 2015 as doi:10.1177/0284185115617349

Original Article

Diagnostic performance of CT, MRI and PET/CT in patients with suspected colorectal liver metastases: the superiority of MRI

Acta Radiologica 0(0) 1–9 ! The Foundation Acta Radiologica 2015 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0284185115617349 acr.sagepub.com

Anselm Schulz1, Ellen Viktil1, Johannes Clemens Godt1, Cathrine K Johansen1, Johann Baptist Dormagen1, Jon Erik Holtedahl2, Knut Jørgen Labori3, Tore Bach-Gansmo1 and Nils-Einar Kløw1

Abstract Background: Meticulous imaging of colorectal liver metastases (CRLM) is mandatory to optimize outcome after liver resection. However, the detection of CRLM is still challenging. Purpose: To evaluate prospectively if magnetic resonance imaging (MRI) with diffusion-weighted and Gd-EOB-DTPAenhanced sequences had a better diagnostic performance for CRLM compared to computed tomography (CT) and fluorine-18 fluorodeoxyglucose positron emission tomography (PET/CT). Material and Methods: Forty-six patients scheduled for resection of suspected CRLM were evaluated prospectively from September 2011 to January 2013. None of the patients had undergone previous treatment for their CRLM. Multiphase CT, liver MRI with diffusion-weighted and dynamic Gd-EOB-DTPA-enhanced sequences and low-dose PET/CT were performed. Two independent, blinded readers evaluated the examinations. The reference standard was histopathological confirmation (81/140 CRLM) or follow-up. Results: A total of 140 CRLM and 196 benign lesions were identified. On a per-lesion basis, MRI had the significantly highest sensitivity overall and for CRLM < 10 mm (P < 0.001). Overall sensitivity/specificity and PPV/NPV were 68%/94% and 89%/81% for CT, 90%/87% and 82%/93% for MRI, and 61%/99% and 97%/78% for PET/CT. For CRLM < 10 mm it was 16%/96% and 54%/80% for CT, 74%/88% and 64%/93% for MRI, and 9%/98% and 57%/79% for PET/CT. Conclusion: MRI had the significantly highest sensitivity compared with CT and PET/CT, particularly for CRLM < 10 mm. Therefore, detection of CRLM should be based on MRI.

Keywords Abdomen/GI, computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET), liver, metastases Date received: 20 March 2015; accepted: 21 October 2015

Introduction Patients with colorectal cancer present in approximately 15–25% cases with colorectal liver metastases (CRLM) at the time of primary diagnosis and a further 15–20% of patients will eventually develop liver metastases after resection of the primary tumor (1). Of patients with synchronous CRLM, up to 77% have metastases confined to the liver, which improves the chance for cure (1,2). Currently, liver resection is the only treatment with documented long-term survival (3–5).

1 Department of Radiology and Nuclear Medicine, Oslo University Hospital, Oslo, Norway 2 The Intervention Centre, Oslo University Hospital, Oslo, Norway 3 Department of Hepato-Pancreato-Biliary Surgery, Oslo University Hospital, Oslo, Norway

Corresponding author: Anselm Schulz, Oslo University Hospital; Department of Radiology and Nuclear Medicine, Ulleva˚l sykehus, Postboks 4956 Nydalen, 0424 Oslo, Norway. Email: [email protected]

Downloaded from acr.sagepub.com at CHINESE UNIV HONG KONG LIB on December 2, 2015

2

Acta Radiologica 0(0)

Meticulous imaging of intrahepatic and extrahepatic tumor deposits is mandatory to achieve optimal surgical results without tumor residuals. Currently, the assessment of CRLM is mainly based on multi-detector row computed tomography (CT), magnetic resonance imaging (MRI), and fluorine-18 fluorodeoxyglucose positron emission tomography fused with low-dose CT (PET/CT) (6). However, the detection of CRLM is still challenging and the optimal diagnostic approach, particularly for lesions of 2 (13), incapability to carry out breath-hold instructions, previous liver resection, other malignancy, pregnancy, contrast media intolerances, reduced renal function, severe claustrophobia, and, in case of metachronous CRLM, 3.0 cm in diameter) or synchronous CRLM.

CT CT scans of the liver were performed on a 40-slice PET/ CT system (Biograph mCT, Siemens AG, Erlangen, Germany). Images were acquired in precontrast, arterial, portal-venous, and late phase after 5 min. The contrast media (150 mL Imeron, 350 mgL/mL, Bracco, Milano, Italy) was administered by a power injector (Stellant, Medrad, Warrendale, PA, USA) via a cubital vein at 5 mL/s. The technical parameters were effective tube current 190 mAs for precontrast, venous phase, delayed phase after 5 min, and 210 mAs for arterial phase, 120 kV tube voltage, 16  1.2 mm collimation, 3.0/3.0 mm slice thickness/increment, 0.6 pitch, 0.5 s rotation time, and 500/500 mm field of view (FOV). Parameters for the low-dose CT for the PET/CT were effective tube current 50 mAs, 1.5 pitch, and 0.33 s rotation time.

MRI Examinations were performed on 1.5T MRI (Achieva, Philips, Eindhoven, The Netherlands) with either SENSE cardiac 32-channels coil or SENSE XL Torso 16-channels. Sequences were axial/coronal T2-weighted (T2W) single-shot (repetition time [TR]/echo time [TE], –/80 ms; FOV, 350.0/350.0 mm; matrix, 336  287; flip angle, 90 ; bandwidth, 337.7 Hz; slice thickness, 5 mm), T1-weighted (T1W) fast field echo (FFE) (TR/TE, 180/ 2.3 ms and 180/4.6 ms; FOV, 350.0/278.6 mm; matrix, 208  148; flip angle, 80 ; bandwidth, 523.3 Hz; slice thickness, 5 mm), diffusion-weighted (b-value 0, 50, 800 s/mm2) (TR/TE, 1313/65 ms; FOV, 376.0/ 303.2 mm matrix, 148  115; flip angle, 90 ; bandwidth, 2705.6 Hz; slice thickness, 5 mm), and T2W turbo-spinecho and spectral presaturation with inversion recovery (SPIR) (TR/TE, 1747/100 ms; FOV, 350.0/278.4 mm; matrix, 444  279; flip angle, 90 ; bandwidth, 252.5 Hz; slice thickness, 5 mm). The patients received 1 mL Gd-EOB-DTPA (Primovist 0.25 mmol/mL, Bayer Schering Pharma AG, Berlin, Germany) per 10 kg bodyweight. The contrast media was injected in a cephalic vein at an injection speed of 1 mL/s by a power injector (Spectris Solaris EP, Medrad, Warrendale,

Downloaded from acr.sagepub.com at CHINESE UNIV HONG KONG LIB on December 2, 2015

Schulz et al.

3

Meeting inclusion criteria n=95 Not included

Random logistical reasons (e.g. absence of research personnel)

n=39

Included into the study n=56 Exclusion

Exclusion

Exclusion

Exclusion

n=1

No diagnostic contrast-enhanced CT

n=5

No MRI

n=3

Withdrawal of informed consent

n=1

No reference standard

Final study population with completed CT, MRI and PET/CT n=46

Inoperable

Resection n 4 26 0

Patients CRLM Histological confirmed CRLM Missed/false negative*: Follow up/histopathology

2/0

Patients CRLM Histological confirmed CRLM Missed/false negative*: Follow up/histopathology

No CRLM n 38 114 81 8/3

Patients** CRLM Histological confirmed CRLM Missed/false negative*: Follow up/histopathology

n 4 0 0 0/0

* Missed by all modalities and identified on follow up or by histopathology ** One patient was resected and histology identified a sclerotic hemangioma Colorectal liver metastases (CRLM)

Fig. 1. Flowchart.

Table 1. Initial patient staging based on the preliminary findings from referring institutions according to AJCC 7th edition. The primary tumor was located in the rectum in 15 patients and colon in 31 patients. Stage

Frequency n

Percent %

Tubulovillous adenoma I IIA IIIB IIIC IVA IVB Total

1 3 7 6 5 22 2 46

2.2 6.5 15.2 13.0 10.9 47.8 4.3 100.0

PA, USA), followed by 20 mL saline flush at 2 mL/s (bolus-tracking). Axial T1W 3D FFE THRIVE images were obtained before contrast and after 20– 30 s, 70 s, 180 s, and 20 min in hepatocyte-phase with additional coronal images (TR/TE, 3.8/1.78 ms; FOV, 350.0/276.3 mm; matrix, 160  138; flip angle, 10 ; bandwidth, 440.1 Hz; slice thickness, 4 mm). Contrastenhanced images were obtained in breath-hold while navigator was used for all other sequences.

PET/CT PET/CT imaging was performed using a scanner with time-of-flight capabilities (Biograph mCT, Siemens AG, Erlangen, Germany). The injected activity was 4 MBq fluorine-18 fluorodeoxyglucose

Downloaded from acr.sagepub.com at CHINESE UNIV HONG KONG LIB on December 2, 2015

4

Acta Radiologica 0(0)

(18F-FDG)/kg/body weight. A whole-body scan from skull base to thigh was performed after 60 min at 2 min per bed position. An axial FOV of 21.6 cm and a 200  200 matrix was used resulting with a pixel size of 4.07 mm. Images were reconstructed by TrueX with time-of-flight (2 iterations, 21 subsets, and Gauss-filtration of 2 mm) with 3 mm slice-thickness. Low-dose CT was used for anatomical correlation and attenuation correction. CT and PET/CT were performed in the same session. If possible, MRI was either performed 6 h after PET/CT on the same day or on the next day. The median time between PET/CT and MRI was 1 day (range, 0–31 days; 83% within 2 days); two patients underwent MRI 30 and 31 days after for logistical reasons.

Diagnostic criteria On CT, hypovascular CRLM were defined as hypoattenuating lesions with less enhancement than normal liver in portal-venous phase and showed washout/ remained hypoattenuating on delayed phase. On MRI, CRLM were defined as hypointense lesions on T1W images, moderate hyperintense lesions on T2W images with poor enhancement centrally and increased peripheral rim enhancement during the early dynamic phases, and hypointense nodules on Gd-EOB-DTPA enhanced MRI images in the hepatobiliary phase (HBP) (10). On DWI, CRLM were defined as hyperintense lesions at b ¼ 50 s/mm2 which remained hyperintense at b ¼ 800 s/mm2 and apparent diffusion coefficient values that were less or equal to that of the adjacent liver parenchyma. Furthermore, an equivalent lesion had to be present on contrast-enhanced images. On PET/CT, CRLM were based on focal FDG uptake exceeding the uptake of the surrounding liver tissue, and distinctly visible on maximal-intensity projection. An uptake was regarded as pathological when the maximum standardized uptake value (SUVmax) within the lesion was significantly higher than the SUVmax of a large ROI within the normal liver, without any numerical cutoff.

Imaging evaluation Four readers with experience in oncological, hepatobiliary, and nuclear medicine imaging were involved in the imaging analysis: one reader (AS) with 7 years of experience, the other three (EV, CKJ, and TBG) each with more than 10 years of experience. Independent evaluation of CT was performed by AS/CKJ, MRI by AS/EV, and PET/CT by AS/TBG. To prevent recognition by AS the minimal time between evaluations was 6 weeks. The readers knew that the patients had colorectal cancer with suspected CRLM, but were

blinded for all other information. Each reader created a report for each exam including number of lesions, characterization, size (maximum axial diameter), and location according to the Couinaud classification. On PET/CT images only CRLM were identified. The two readers agreed to a final consensus.

Reference standard During resection, the liver was inspected, and during open procedures bimanually palpated. Intraoperative ultrasound was used if requested by the surgeon. The specimens were formalin fixated for histopathological examinations and depending on the shape of the specimen cut into parallel slices of 3 mm in axial plane. Hematoxylin and eosin staining were provided. Clinical data, operation reports including the number of resected lesions were available to the pathologist, but a detailed radiological imaging report was usually not given. Unidentified CRLM, when found by the histopathology examination or identified on the first postoperative routine CT control (MRI in 4 cases) after 4 months, were defined as false negative CRLM. For size dependent subgroup analysis those CRLM were considered