Can contrast-enhanced ultrasonography replace multidetector-computed tomography in the detection of liver metastases from colorectal cancer?

Results Three hundred and sixty-five patients were included. All patients had undergone preoperative US, CEUS and MDCT and 65.5% had received IOUS. The gold standard found liver metastases in 54 patients (14.8%). Multidetector CT found significantly more metastases than CEUS in 15 (28%) of the patients ( p = 0.02). In a patient-by-patient analysis MDCT had a non-significantly higher sensitivity in the detection of liver metastases compared to CEUS (0.89 versus 0.80, p = 0.06). The specificity of CEUS (0.98) was slightly better than that of MDCT (0.94) ( p = 0.02). Conclusion Multidetector CT found significant more metastases than CEUS, and MDCT had in patient-by-patient analysis a non-significant better sensitivity ( p = 0.06) in detecting liver metastases in patients with CRC. Keywords Multidetector computer tomography Ultrasound contrast agent Liver neoplasms Metastases Colon cancer Rectal cancer 1 Introduction Colorectal cancer (CRC) is the second most common cause of death due to cancer in the Western world, with a 5-year survival rate of 55%. The liver is the most common site for metastases from CRC, which are the main cause of death. In patients with CRC, 15–25% have synchronous liver metastases, and a similar number of patients develop liver metastases after colorectal resection with curative intent [1] . The presence of hepatic metastases and the extent of liver involvement have a great impact on the treatment and prognosis for patients with CRC. Several imaging techniques are available for preoperative staging of liver metastases. In clinical practices, MDCT, magnetic resonance imaging (MRI) and US are the most accepted diagnostic techniques worldwide for the detection of liver metastases from CRC, while CT during arterial portography and intraoperative ultrasonography (IOUS) improve sensitivity and specificity and are used as preoperative assessments of surgical candidates for liver resection [2,3] . The development of CEUS has increased the potential of US in the assessment of liver metastases. The sensitivity of assessments using conventional transabdominal ultrasonography (US) is relatively low (53–77%) [4] , but CEUS seems to improve sensitivity in detecting liver metastases by 15–50% [5–7] ( Fig. 1 ). The reported sensitivity of single-slice CT in the detection of liver metastases from CRC is also low (60–73%) [8] , but MDCT has a better spatial resolution in the direction of the body axis and greater anatomic coverage during a single breath hold [9] . MDCT performed with thinner section thickness (i.e., 2.5 or 5 mm) seems to improve the sensitivity of liver metastasis detection [10] . The purpose of this study was to investigate if CEUS can replace MDCT in the preoperative detection of liver metastases in a group of unselected patients with CRC by comparing the sensitivity and specificity of the two techniques. 2 Materials and methods 2.1 Patient population Four hundred and sixty-one consecutive patients were referred to and evaluated at the Department of Colorectal Surgery, Aarhus University Hospital with known or suspected primary or local recurrence of CRC between June 2003 and May 2005. Forty-five patients did not meet the inclusion criteria because they were pregnant or breast-feeding woman or because they had one of the following conditions: a known allergy to X-ray contrast or sulphur hexafluoride, renal insufficiency ( s -creatinin >200 μmol/l), severe heart failure, known right–left shunts, serious pulmonary hypertension (pulmonary pressure >90 mm Hg), unmanageable systematic hypertension and ARDS, tumour from a known malignancy other than colorectal neoplasia (e.g., sarcoma or lymphoma). Another 51 patients were not included for logistic reasons (mainly due to lack of available members of the research team). All of these patients had performed US and MDCT and according to the results of these examinations six patients (11.7%) had liver metastases. Finally, 365 patients (161 women and 204 men, with an age range of 23–93 years and a mean age 67 years) were included in the study, which was approved by the local ethical committee. Two hundred and seventy-four (75%) were patients with primary CRC, 57 (16%) were patients with local recurrence of CRC referred to the surgical department for curative treatment, 5 (1%) were patients with other cancers, and 29 patients (8%) had tumour specimens showing benign tumours. Ultrasonography and CEUS were performed first, and biphasic MDCT scanning was performed within 2 days ( n = 365). An intraoperative ultrasonography ( n = 239) was performed, when the patients were operated ( n = 296), within 0–19 weeks (mean of 23 days) of these initial procedures. One hundred and twenty-six patients did not had an IOUS because of patient not operated ( n = 69), logistical and technical reason when performing IOUS (for example emergency operation because of ileus, technical not possibly to scan liver because of adherence) ( n = 43) and other reasons (other operation procedure, like transrectal endoscopic microsurgery (TEM), bleeding and low blood pressure under the operation) ( n = 14). Ultrasonography-guided fine-needle biopsies ( n = 47) were performed on all patients with suspicious lesions after the MDCT had been completed. Only in case of technical not possibly or of ethical reasons, fine-needle biopsy was not performed. As criteria for sufficient biopsy, the tip of the fine-needle clearly has to be recognized in the lesions, when performing the aspiration by the fine-needle. For all imaging modalities, the findings were immediately documented according to a standard protocol, including the number, size and localization of the liver lesions. Couinaud's classification of liversegmental anatomy was used. Up to five metastases were observed in every patient in the study. In patients with more than five metastases, patients were grouped into those with 6–10 metastases and those with more than 10. Benign lesions were also registered. If any inconsistency was found in the results based on gold standard, a control MDCT, CEUS and, when indicated, a fine-needle biopsy, were performed 3 months later. If fine-needle biopsy was not possible, any progression of size or number of lesions was classified as metastases. 2.2 Sonographic examination technique All CEUS were performed by one of three radiologists with between 7 and 25 years of US experience. For consistency in imaging techniques and analyses, all examinations were performed with the same system (EUB 8500, Hitachi (CFM)) and used a curved 3.5 MHz transducer and phase inversion technique. A low mechanical index (MI) of 0.08–0.18 (mean 0.10) was selected. The CEUS was performed after administration of 2.4–5.0 ml of 8 μl per milliliter Sonovue (Bracco, Italy) through the brachial vein as a bolus followed by a flush with 5-ml isotonic saline. The delay before the scan was performed was between 45 and 90 s after this administration. The time of the infusion was reported for each CEUS. No patients were excluded because of obesity or other difficult scanning conditions. Before CEUS, an US examination was performed to assess the morphology of the lesions. Then the CEUS was done and metastases were defined as round, oval and lobulated hypoechoic focal defects in a contrast-enhanced liver parenchyma in the portal venous or delayed vascular phase. The CEUS examinations were recorded on CD-ROM and were evaluated at the time of examination. The intraoperative ultrasonography was performed using a commercially available US system (EUB 6000, Hitachi) with a 5.0–10.0 MHz transducer (T-shaped probe) as a part of the routine procedure in CRC surgery. A surgeon and one of the ultrasonography investigators, who did not perform the CEUS, completed the procedure. The IOUS was first performed without any knowledge of the results of the preoperative examinations. The “code was cracked” and the results of the CEUS and MDCT were evaluated. If preoperative examinations showed metastases which were not detected by the initial IOUS, IOUS and palpation were repeated. 2.3 MDCT technique The MDCT scans were obtained using a multidetector row helical CT system (Phillips Marconi Mx 8000, 4-slice), with the following parameters: a 2.5 mm collimation, an increment of 1.6, a pitch factor of 1.25, a table speed of 0.75/s and 120 kV with a maximum current of 150 mA/slice. Arterial-phase and portal venous-phase images were obtained beginning 30 and 70 s, respectively, after the start of an injection of 100 ml non-ionic contrast medium (Visipaque 270 mg iodine per milliliter (Amersham Health)). This injection was administered at a rate of 4 ml/s with a power injector guided by the bolus-chase technique with a threshold of 150 HU. All examinations by MDCT were performed after the CEUS. One of two investigators, blind to the results of the CEUS, evaluated the images, and the number, size and location of metastases were documented. 2.4 Gold standard The gold standard was the number of metastases revealed by the combination of US, CEUS, MDCT, IOUS, fine-needle biopsy, palpation and repeated IOUS. If there were any inconsistencies in results, additional US, CEUS, MDCT and, if relevant, fine-needle biopsy were performed 3 months postoperatively. If a liver resection was performed, the pathological examination of the resected specimen would contributed to gold standard. 3 Statistical methods The sensitivity and specificity of CEUS and MDCT in the assessment of liver metastases were calculated using the results of gold standard as the standard. A comparison of sensitivity and specificity in this patient-by-patient analysis, was performed using McNemar's test. Ninety-five percent confidence intervals (CI) for overall sensitivity and specificity values were calculated using the binomial distribution. The Wilcoxon matched-pairs signed-rank test was used to compare the numbers of detected metastases by MDCT and CEUS, and an unpaired Student's test was used to compare the average size of liver metastases. For all analyses, p < 0.05 was considered to indicate a statistically significant difference. 4 Results Based on the gold standard, 54 of 365 patients (14.8%) had one or more liver metastases. Forty-seven patients had primary and seven patients had local recurrence of CRC. The prevalence of liver metastases was 17.2 for primary and 12.3% for recurrence of CRC. The final diagnosis of the 54 patients with liver metastases was based on the following: In 43 patients, metastases were identified with both CEUS and MDCT (confirmed by IOUS ( n = 12) or fine-needle biopsy ( n = 35)). Five cases were found only by MDCT (all confirmed by fine-needle biopsy), four cases were found by IOUS (three confirmed by fine-needle biopsy and one by follow-up), and two by intraoperative palpation and fine-needle biopsy (both very small and located on the surface of the liver). Based on results of the gold standard, one metastasis was identified in 13 patients, two metastases in five patients, three in two patients, four in seven patients, five in one patient, 6–10 in 10 patients and, finally, more than 10 metastases were identified in 16 patients. In 26 cases there was a suspicion of liver metastases by either the CEUS or MDCT and/or IOUS examinations, while 3 months follow-ups with a fine-needle biopsy ( n = 6) disproved the suspicions. 4.1 Sensitivity of MDCT versus CEUS Multidetector CT found significantly more metastases than CEUS ( p = 0.02) ( Table 1 ). In a patient-by-patient analysis, MDCT found metastases in 48 of the 54 patients (88.9%, CI [77.3–95.8%]), while CEUS identified metastases in 43 of 54 patients (79.6%, CI [66.5–89.4%]), but the difference (9.3%, CI [−0.3 to 18.8%]) in sensitivity was not significant ( p = 0.063) ( Table 2 ). In 15 (28%) of patients with liver metastases, MDCT found more metastases, in 35 patients, MDCT and CEUS found the same number of metastases, and in four patients CEUS identified more metastases than MDCT (one of the four cases followed up with MDCT 3 months later could not confirm the number of metastases found with CEUS). In the 11 false-negative patients by CEUS, six were identified by IOUS and palpation performed an average of 26 days (15–35 days) after the CEUS, and five patients were found by MDCT (four of the five patients had one metastasis, one had three metastases). Seven patients had one metastasis, two patients had two metastases, one had three metastases and one patient had more than five metastases. The average size of liver lesions missed and detected by CEUS was, respectively, 1.15 cm (range 0.8–2.5) and 1.67 cm (range 0.4–4.5). The mean numbers of metastases per patient detected by CEUS was 6.11. Of the six false-negative patients on MDCT, none were detected by CEUS. Using reference imaging, three patients were found to have one metastasis, two patients had two metastases and one had more than five metastases (found by US 1-month later and confirmed by autopsy 2 months later). The average size of missed and detected liver lesions by MDCT was, respectively, 1.16 cm (range 0.8–1.5) and 1.68 cm (range 0.4–4.5). The mean numbers of metastases per patient detected by MDCT was 5.83. 4.2 Specificity of MDCT versus CEUS According to gold standard 311 patients did not have liver metastases. Contrast-enhanced ultrasonography detected 304 of these patients, while MDCT detected 293. Therefore the specificity of CEUS was 97.7% (304/311) (CI: [95.4–99.1%]), while that of MDCT was 94.2% (293/311) (CI: [91.0–96.5%]). The difference in specificity, 3.5%, CI: [0.5–6.5%]) was significant ( p = 0.02) ( Table 2 ). 5 Discussion Multidetector CT found significantly more metastases compared with CEUS. In a patient-by-patient analysis, sensitivity of MDCT and CEUS was 88.9 and 79.6% ( p = 0.06), respectively, while the specificity was better for CEUS compared to MDCT (respectively 97.7 and 94.2%, p = 0.02). Four of the five patients found by MDCT and not by CEUS had one metastasis and they were candidates for liver resection, and the clinical benefit of using MDCT instead of CEUS seems relevant. The main strength of this study is that a large number of patients have been included, and the study design was consecutive, prospective and double blinded unlike in other studies [11,12] . Secondly, our study design reflects a “true clinical preoperative situation”; we included patients with primary or local recurrence of CRC without known liver metastases. We did not exclude patients with other cancers, and patients who had tumour specimens showing benign tumours, because the study population should represent a “true clinical preoperative situation” in patients who were referred to and evaluated on a department of colorectal surgery with known or suspected primary or local recurrence of CRC. The overall prevalence of liver metastases in patients with CRC was 14.8% in our study. Other studies have used a population with a very high prevalence (63–65%) of liver metastases, or have included patients with established metastases [5,6] . This selection may lead to a greater accuracy [4] . Thirdly, we compared the sensitivity on a patient-by-patient analysis and did not use a lesion-by-lesion analysis, like earlier studies [13] . A lesion-by-lesion analysis in patients with many liver lesions is a potential source of error [13] , while it is well known, that with many liver lesions it can be difficult to compare the same lesions using two different imaging modalities (in our study: CEUS and MDCT), despite the assignment of lesions to their Couinaud segment. In our study, the sensitivity of MDCT was slightly better than CEUS, which is not in accordance with two other studies based on a patient-by-patient analysis, which both included a study population with a remarkably high prevalence of liver metastases [5,14] . In the study of Dietrich et al., the sensitivity of CEUS was slightly better than MDCT (both were remarkably high, with values of 91.2 and 89.2%, respectively). Konopke et al. also found that the sensitivity of CEUS was better compared with MDCT (86.0 and 76.0%, respectively), but the difference was not statistically significant and the protocols of how and when the MDCT was performed are unclear. The prevalence of liver metastases in patients with primary CRC (17.2%) in our study is in accordance with other studies [4,15] . There were also important limitations in our study. Of logistic reasons we did not include 51 patients, which is a important potential bias for a prospective study. Whether this selection had any influence on the results or not, is difficult to predict. According to US and MDCT 6 of the 51 patients had liver metastases, all seen both by US and MDCT. The MDCT was performed according to the existing protocol from our department, but the volume of 100 ml contrast with an iodine concentration of 270 mg/ml administered independently of patient weight may have resulted in a significantly low total amount of injected iodine, and clearly less than used in several other studies [16,17] . Moreover, we used a 4-slice MDCT, but at present many institutions use 16- or 64-slice MDCT which are known to have a better sensitivity of liver metastases detection [10] . It is likely that a MDCT performed with an optimized protocol would favor the observed differences. On the other hand, the experience and technique of CEUS have improved since this study was conducted, which may influence the results of this study. However, CEUS still has limitations. It is observer dependent, and in obese patients, the accuracy in some parts of the liver is limited because of reduced penetration and a more inherent problem with the documentation and management of images. In longitudinal studied patients, where examinations have to be compared, long recorded video sequences from CEUS is more difficult to handle than are still images from MDCT. We may have overestimated the sensitivity of both CEUS and MDCT, because the reference imaging had included IOUS in only 66% of the patients. However, this bias would likely not influence the significant difference in the sensitivity of liver metastases by CEUS and MDCT. On the other hand, the time difference between CEUS/MDCT and IOUS (a mean of 23 days) may result in an underestimation of the sensitivity of CEUS/MDCT because of the risk of developing metastases over time. We performed the IOUS without total surgical mobilization of the liver, which may lead to a suboptimal visualisation of all liver segments ( Fig. 2 ). Finally, in 7 out of 54 cases, fine-needle biopsy of liver lesions was not performed, largely due to technical reasons. However, in four of these cases the lesions were confirmed by follow-up at the Department of Oncology. It is controversial whether biphasic technique in MDCT for liver metastasis detection from CRC improves sensitivity. One study [18] has suggested that some liver metastases missed in the portal phase were visible in the arterial phase. Arterial-phase imaging increases the detection of hypervascular tumours such as hepatocellular carcinaoma or hypervascular metastases [19] . However, most liver metastases from CRC are hypovasculare and, therefore, are best imaged in the portal phase. Authorities [19,20] agree that an arterial phase is not necessary for MDCT of metastases from CRC. Therefore, the biphasic technique would probably not influence the sensitivity, but may improve the specificity of the MDCT. When comparing the accuracy of two imaging modalities in detecting liver metastases, the problem will always be in defining the gold standard. We used a combination of US, CEUS, MDCT, IOUS, fine-needle biopsy, palpation and repeated IOUS. To further improve the gold standard, we made a follow-up 3-month later if any inconsistency was found in the results of the primary examinations. 6 Conclusion CEUS cannot replace MDCT in liver metastases screening in patients with CRC. Multidetector CT detects more metastases and the sensitivity of this technique tends to be higher ( p = 0.06). In addition, with MDCT it is easier to compare repeated examination in patients treated for CRC. Acknowledgements Financial support for this study was provided by The Danish Cancer Society and Aarhus County. References [1] D. Tong A.H. Russell L.E. Dawson W. Wisbeck Second laparotomy for proximal colon cancer. Sites of recurrence and implications for adjuvant therapy Am J Surg 145 1983 382 386 [2] P. Soyer M. Levesque D. Elias G. Zeitoun A. Roche Detection of liver metastases from colorectal cancer: comparison of intraoperative US and CT during arterial portography Radiology 183 1992 541 544 [3] Y.K. Kim S.W. Ko S.B. Hwang C.S. Kim H.C. Yu Detection and characterization of liver metastases: 16-slice multidetector computed tomography versus superparamagnetic iron oxide-enhanced magnetic resonance imaging Eur Radiol 16 2006 1337 1345 [4] C. Glover P. Douse P. Kane Accuracy of investigations for asymptomatic colorectal liver metastases Dis Colon Rectum 45 2002 476 484 [5] R. Konopke S. Kersting H.D. Saeger A. Bunk Detection of liver lesions by contrast-enhanced ultrasound—comparison to intraoperative findings Ultraschall Med 26 2005 107 113 [6] T. Albrecht M.J. Blomley P.N. Burns Improved detection of hepatic metastases with pulse-inversion US during the liver-specific phase of SHU 508A: multicenter study Radiology 227 2003 361 370 [7] L.P. Larsen M. Rosenkilde H. Christensen The value of contrast enhanced ultrasonography in detection of liver metastases from colorectal cancer: a prospective double-blinded study Eur J Radiol 62 2007 302 307 [8] K.W. Kim A.Y. Kim T.K. Kim Small (更多