Difference in the performance of diagnosing colorectal liver metastasis between computed tomography and carcinoembryonic antigen

Background: The study was performed to compare the diagnostic roles of computed tomography (CT) and carcinoembryonic antigen (CEA) in colorectal liver metastasis (CRLM). Methods: 255 patients with colorectal cancer (CRC) were enrolled. These patients were conrmed as CRLM by histopathological assay. CT scans of the liver were performed with a 64-slice CT system. Serum CEA levels were determined using a human circulating cancer biomarker magnetic bead panel. True positive (TP), false positive (FP), true negative (TN) and false negative (FN) were calculated for CT and CEA with histopathological assay as golden standard. Results: 142 CRLM patients and 113 non-CRLM patients were conrmed in the study. There were no obvious differences in age, sex and Dukes stage between CRLM and non-CRLM patients (P (cid:0) 0.05). Diagnostic roles of CT and CEA on per-patient and per-lesion were analyzed. Detection sensitivity, specicity, positive predictive value (PPV) and negative predictive value (NPV) of CT on per-patient basis were 74.7%, 56.4%, 52.1%, and 77.9%, respectively. Sensitivity, specicity, PPV and NPV of CEA were 64.4%, 51.4%, 49.3%, and 67.3%. In the analysis on per-lesion basis, detection sensitivity, specicity, PPV, and NPV of CT were increased to 87.8%, 88.8%, 91.5%, and 84.1%, respectively. Detection sensitivity, specicity, PPV and NPV of CEA on per-lesion basis were 82.6%, 64.4%, 63.4%, and 83.2%. Conclusion: CT exhibited better performances than CEA in diagnosis of CRLM on both of per-patient and per-lesion basis.


Background
Colorectal cancer (CRC) often metastasizes to other organs, among which liver is the most common site [1]. Fifteen to twenty-ve percent cases with colorectal liver metastases (CRLM) are presented among CRC patients at primary diagnosis. About 15%-20% of CRC patients will develop liver metastases after resection surgery on primary tumor [2]. Detection of CRLM has important in uence on treatment methods and prognosis of CRC patients. Speci c information about the localization, number and size of metastasized liver lesions contributes to making decision on surgery, adjuvant therapy or palliative therapy [3,4].
Tumor visualization is commonly performed with imaging techniques such as magnetic resonance imaging (MRI), ultrasonography and computed tomography (CT) [5]. Albrecht et al. (2014) reported that CT sensitivity and speci city were 63.55%-89.71% and 78.60%-94.41%, respectively. It also demonstrated that CT interpretation in detection of CRLM is less affected compared to MRI analysis [6]. A metaanalysis demonstrated that diagnostic sensitivity of CT was 74.4% and 83.6% on a per-patient basis and per-lesion basis, respectively. The detection accuracy of CT was lower than MRI and positron emission tomography (PET) [7].
Carcinoembryonic antigen (CEA), rst described in 1965, is one of intracellular glycoprotein observed in 90% of CRC cases [8]. It serves as intercellular adhesion molecules that accelerate the aggregation of CRC cells [9,10]. CEA, also named as CD66a and CEACAM5, is one of diagnostic biomarkers for many cancers [11][12][13][14][15]. For CRC and CRLM, CEA is the most commonly used biomarker. It is primarily applied to test recurrence of CRC rather than screening because of low sensitivity and speci city [16,17]. However, this marker molecule is much more sensitive in diagnosis of lymph node metastases and CRLM than lung metastases and local recurrence [16][17][18].
Our study aimed to evaluate the diagnostic roles of CT and CEA in CRLM and compared their differences in diagnostic sensitivity, speci city, positive predictive value (PPV) and negative predictive value (NPV).

Methods
This study was approved by the ethics committee of Southwest Hospital, Army Medical University. Written informed consent was obtained from all patients.
A total of 255 CRC patients from the hospital were enrolled. They were con rmed as primary CRC by histopathological assay. Among them, 142 patients suffered liver metastases and 113 patients did not. CRLM patients, synchronous or metachronous, were diagnosed by histopathological assay. The individuals with extra-hepatic metastasis were excluded from the study. The basic information of patients was showed in Table 1.    [19,20]. The detection of liver metastases occurred in CRC patients is crucial for treatment strategy and high detection sensitivity is also important.
Unenhanced ultrasound (US) technology on the liver shows a detection sensitivity of 50%-70% for synchronous liver metastases [21,22]. However, US possess a limitation of low imaging contrast between liver parenchyma and liver lesions. Bipat et al. suggested that US should simply be used to differentiate patients with diffuse CRLM from patients with few metastases [23]. As for CRLM, the information about liver metastases, such as number, size, distribution location and volume of the lesions as well as volume of the rest should be grasped in detail before selecting surgical treatment [24]. Fortunately, CT is able to exactly complete this task. Until now, CT is the main diagnosis method for CRLM due to its highresolution imaging and re-modi ed images, which makes the inspection of small metastases available.
Meanwhile, the accurate measurement of volume and the location description of metastases by portal venous anatomy and imaging hepatic arterial about normal and tumor liver by CT are both equally important for planning the surgical resection of CRLM [24]. The diagnostic sensitivity of contrast enhanced CT in liver metastases ranges from 68-85% [25][26][27][28].
In addition of these imaging modalities, tumor markers of type collagen (COLIV) and CEA were also used in the diagnosis of CRLM. Nyström et al. measured COLIV and CEA levels in CRLM patients and found that circulating COLIV level increased in 81% and CEA in 56% of CRLM patients at diagnosis. It also demonstrated that the diagnostic sensitivity of COLIV was lower than CEA [29]. Besides, Zhang et al. reported that serum CEA level was an independent prognostic predictor for CRLM [30]. Positron emission tomography (PET)/computed tomography (CT) was demonstrated to be superior to CEA in diagnosis of CRC and its recurrences [31].
Our study was performed to compare the diagnostic roles of CT and CEA in CRLM. On per-patient basis, CT showed better performance than CEA, exhibiting in sensitivity, speci city, PPV and NPV. On per-lesion basis, the diagnostic indexes of CT and CEA showed certain improvements, however, CT was still superior to CEA. The study rstly evaluated the difference in diagnostic accuracy between CT and CEA and the results were reliable and credible. However, several points must be noted in the analysis. Only 255 CRC patients were enrolled and the sample size was relatively small, hence, future larger-scale studies should be conducted to con rm the results. Moreover, more diagnostic modalities, such as MRI, PET/CT and contrast enhanced ultrasound (CEUS) should be considered, which will provide more accurate results about diagnosis methods of CRLM.

Conclusions
In a conclusion, CT was superior to CEA in diagnosis of CRLM. The results may contribute to the improvement in early diagnosis of CRLM and selection of surgical methods. The subjects had been informed the objective. Certainly, written consents were signed by every subject in this study.

Consent for publication
We obtaining permission from participants to publish their data.

Availability of data and materials
All data generated or analysed during this study are included in this published article.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions K.Z. and Z.W. design of the work; T.M. and Z.H. the acquisition, analysis, X.W. and Y.P. interpretation of data; Y.C. and Y.D. the creation of new software used in the work; Z.R. have drafted the work or substantively revised it. All authors read and approved the nal manuscript.