Abstract
Background:
Adopting an appropriate noninvasive radiological method is crucial for periodic surveillance of liver metastases in colorectal cancer (CRC) patients after surgery, which is closely related to clinical management and prognosis. This study aimed to prospectively enroll stage II-III CRC patients for the surveillance of liver metastases and compare the diagnostic performance of contrast-enhanced CT (CE-CT) and non-enhanced abbreviated MRI (NE-AMRI) during this process.
Methods:
587 CRC patients undergoing radical resection of the primary tumor were evaluated by 1 to 3 rounds of surveillance tests, consisting of abdominal CE-CT and contrast-enhanced MRI (CE-MRI) within 7 days at 6-month intervals. Subsequently, images of NE-AMRI were extracted from the CE-MRI examination, and paired CE-CT and NE-AMRI analysis were performed. The lesion-based detection rates between two protocols were compared, and a subgroup analysis was performed in lesions with a size of ≤10 mm. The patient-based sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and the areas under the curves (AUCs) of CE-CT and NE-AMRI in each round were evaluated. Finally, the relationship between the diagnostic accuracy of two protocols and characteristics of patients was explored.
Results:
The lesion-based detection rates of NE-AMRI in three rounds were all significantly higher than those of CE-CT (P < 0.001, P < 0.001, P = 0.003, respectively). In the subgroup analysis of lesions ≤ 10 mm, NE-AMRI also performed better than CE-CT (P < 0.001, P = 0.002, P = 0.005, respectively). The patient-based sensitivities, specificities, NPVs, and PPVs of NE-AMRI were higher than those of CE-CT in three rounds of surveillance. The AUCs for NE-AMRI were all significantly better than those for CE-CT in each round (P = 0.015, P = 0.045, P = 0.009, respectively). Furthermore, patient BMI and fatty liver disease had impacts on the diagnostic accuracy of the CE-CT protocol, but not on the NE-AMRI protocol.
Conclusion:
NE-AMRI may be a promising periodic surveillance tool for CRC patients after surgery to increase diagnostic accuracy of liver metastases, developing personalized clinical management and improving prognosis, simultaneously avoiding side effects associated with ionizing radiation and contrast agents.
Keywords: colorectal cancer, contrast-enhanced CT, liver metastases, non-enhanced abbreviated MRI, periodic surveillance after surgery
Introduction
Highlights.
NE-AMRI showed significantly better diagnostic performance than CE-CT as a surveillance tool for liver metastases in three rounds of evaluation for 587 CRC patients undergoing radical resection of the primary tumor (P = 0.015, 0.045, 0.009, respectively).
The diagnostic accuracy of NE-AMRI protocol did not change with patient age, gender, and BMI, but patient BMI and fatty liver disease had significant impacts on the diagnostic accuracy of CE-CT protocol.
Colorectal cancer (CRC) is currently the third most commonly cancer, and the liver is the main target organ for metastases of CRC[1,2]. Approximately 15% to 25% of patients are synchronously diagnosed with colorectal liver metastases (CRLM), while another 15% to 25% of patients will develop liver metastases after the resection of primary tumors[3]. Routine surveillance imaging may allow an earlier detection of CRLM, which could facilitate timely initiation of hepatic resection or systemic therapy before the disease progresses beyond cure[4]. Therefore, surveillance imaging for the occurrence of CRLM is now frequently performed in CRC patients after surgery, especially in patients with stage II-III disease, who are at higher risk of getting liver metastases[5].
International guidelines (ESMO, NCCN, and JSCCR) recommend that CRC patients should undergo abdominal contrast-enhanced computed tomography (CE-CT) scans at regular intervals of 6–12 months after surgery for up to 3–5 years[6]. Recent studies, however, revealed that CE-CT had low sensitivity for detecting CRLM lesions, particularly for lesions smaller than 10 mm[7,8]. Because of the repetitive nature of surveillance scans, the cumulative radiation hazard and of periodic CT should not be neglected[9]. Contrast-enhanced magnetic resonance imaging (CE-MRI) has shown excellent performance in detecting liver metastases, suggesting that it may be the alternative method for CRLM surveillance[10]. However, the high cost and long imaging acquisition time of periodic full-protocol CE-MRI hindered its widespread use. CE-MRI also has various drawbacks associated with the use of gadolinium agents, including anaphylaxis and long-term retention in human tissues[11,12]. The Royal College of Radiologists UK published their position statement emphasizing that gadolinium agents should only be used when essential diagnostic information cannot be obtained with unenhanced scans[13]. The non-enhanced abbreviated MRI (NE-AMRI) protocol, only including T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), diffusion-weighted imaging (DWI), and apparent diffusion coefficient (ADC) maps, has recently seemed to be a promising option for evaluation of focal liver lesions[14,15]. Hwang et al. reported that NE-AMRI had the comparable value for characterizing liver metastases to gadoxetic acid-enhanced MRI, which may serve as an alternative for the detection of CRLM[16]. However, a few studies compared the diagnostic performance of CE-CT and NE-AMRI for the surveillance of CRLM.
In this study, we aimed to evaluate and compare the diagnostic performance of CE-CT and NE-AMRI protocols in the surveillance of liver metastases in a cohort of stage II-III CRC patients within 18 months follow-up after surgery.
Methods
Study population
This prospective observational study was approved by the institutional review board of our hospital (No. Y2020-120), written consent was obtained from all study participants, and all data were confidential. The reported work was in line with the STROCSS criteria[17]. Patients were recruited from outpatient clinics, multidisciplinary team meetings, and inpatient wards of the department of colorectal surgery between July 2020 and December 2022. The inclusion criteria were (a) ≥ 18-year-old and ≤ 80-year-old; (b) diagnosed with stage II-III CRC and undergoing radical resection of the primary tumor; (c) Eastern Cooperative Oncology Group performance status ≤ 2; (d) without other primary malignant tumors; and (e) without contraindications for CE-CT and CE-MRI examinations. Finally, 587 CRC patients were enrolled to monitor the occurrence of liver metastases. Patients were evaluated by 1 to 3 rounds of surveillance tests, consisting of abdominal CE-CT and CE-MRI within 7 days at 6-month intervals. Subsequently, images of NE-AMRI were extracted from the CE-MRI examination, and paired CE-CT and NE-AMRI analysis were performed. For every round, the criteria of lost to follow-up were (a) withdrawn from the study or died without subsequent follow-up examinations and (b) unable to cooperate resulting in incomplete inspection or poor image quality. For example, of 587 patients enrolling in the initial surveillance, 15 were lost to follow-up due to the criteria (a) and 13 were lost to follow-up due to the criteria (b). Finally, only 559 patients completed the first round of evaluation. The flow diagram of the study is shown in Figure 1.
Figure 1.
Flowcharts of study participants.
Patient and primary tumor characteristics
The information of initial enrolled patients, such as age, sex, BMI, whether suffer fatty liver disease, CEA, and CA199 levels, was collected. The diagnosis of hepatic steatosis is based on liver Hounsfield unit (HU) difference on CT between the liver and spleen, typically the liver-to-spleen ratio[18]. The formalin-fixed and paraffin-embedded surgical specimens of CRC tumors were collected from the pathology department. Postoperative pathological results of primary tumor location, mucinous histology, T and N stages, vascular invasion, perineural invasion, and KRAS and BRAF mutation status were determined by two pathologists, respectively, with 15 and 19 years of experience in RC pathology. Samples of which two pathologists disagreed on the diagnosis were re-analyzed until a consensus was reached.
Acquisition of CE-CT and CE-MRI images
Abdominal three-phase contrast-enhanced CT scans were performed using multi-slice CT instruments (SOMATOM Sensation 64, SOMATOM Definition AS or SOMATOM Force, Siemens Healthcare, Germany, and Aquilion One 320, Toshiba Medical Systems, Tokyo, Japan). The parameters were as follows: 120 kV tube voltage, 130–192 mA tube current, 3–5 mm slice thickness, and 512 × 512 matrix. About 100 mL nonionic contrast agent (300 mg I/mL; Iohexol 350) was injected into the cubital vein with a flow rate of 3.0 mL/s. Plain-phase images were acquired first; then, contrast-enhanced CT scans were started at 6 seconds (arterial phase) and 46 seconds (venous phase) after a trigger threshold of 100 Hounsfield units was reached at the abdominal aorta.
Abdominal contrast-enhanced MR scans were performed using following instruments: 1.5-T uMR560 scanner (United Imaging Healthcare, Shanghai, China), 3.0-T uMR770 scanner (United Imaging Healthcare, Shanghai, China), Aera 1.5-T Siemens MR (Siemens Healthcare, Germany), and Verio 3.0-T Siemens MR (Siemens Healthcare, Germany). The parameters for the full sequences of different MR scanning systems are listed in Table S1, http://links.lww.com/JS9/D799. Taking the uMR560 scanner as an example, NE-AMRI sequences were T1-weighted (T1W) in-phase and opposed-phase gradient echo sequence, T2-weighted (T2W) fat-suppressed fast spin-echo sequence with breathhold, and diffusion-weighted (DW) single-shot spin-echo planar sequence with b values of 50 and 800 s/mm2. The full-sequence contrast-enhanced MRI additional included dynamic imaging, which utilized a 3D breathhold T1W fat-suppressed gradient-echo sequence. The contrast agent (Gadobutrol; Bayer HealthCare) was administered at a dose of 0.1 mmol/kg at a rate of 2 mL/s. Arterial phase images were obtained 7 seconds after the contrast agent had arrived at the thoracic aorta, and images of the portal venous phase and delayed phase were subsequently obtained 60 and 180 seconds after administration of the contrast agent, respectively.
Image analysis
The images of CE-CT and NE-AMRI were all anonymized. In each round, images were independently analyzed by 2 reviewers (with 8 and 15 years of clinical experience in abdominal radiology, respectively). They reviewed all the images obtained during the CE-CT and NE-AMRI examinations on the basis of the results of some prior examinations, including clinical data and primary tumor characteristics, but they were blinded to the results of reference standards This partially non-blinded setting was selected so that the study results would be obtained in a typical clinical setting. To control reviewers’ recall bias, NE-AMRI images were reviewed 6 weeks after the evaluation of CE-CT images. Interobserver agreement was subsequently evaluated for lesion-based analyses. If discrepancies between the two reviewers were found, images were evaluated by a senior radiologist (with 20 years of clinical experience in abdominal radiology) for further patient-based analyses. The radiological diagnostic criteria of liver metastases on NE-AMRI were (a) moderate low signal intensity on T1WI, (b) mild-moderate high signal intensity on T2WI, (c) higher signal on DWI, and lower or similar signal to adjacent liver on the corresponding ADC map. The criteria for liver metastases on CE-CT were defined as a hypovascular nodule with irregular peripheral enhancement. If the size of the liver lesion was significantly increasing in the process of follow-up, it was also diagnosed as metastases.
Reference standard
The diagnosis of CRLM was established by an independent multi-disciplinary team including two pathologists, two liver surgeons, two radiologists, and one ultrasound expert. To obtain a definite diagnosis, the multidisciplinary team can use and review all patient clinical data available. First, the multidisciplinary team mainly depends on full-sequence CE-MRI and follows up radiological data to support and reject the diagnosis of CRLM. The diagnostic criteria of liver metastases on CE-MRI were (a) moderate low signal intensity on T1WI, (b) mild-moderate high signal intensity on T2WI, (c) higher signal on DWI, and lower or similar signal to the adjacent liver on the corresponding ADC map, and (d) irregular peripheral enhancement during the arterial phase or portal venous phase. And if the follow-up radiological data demonstrated a new lesion or recently enlarged lesion, the lesion was consider to be positive. Notably, the evaluators of CE-CT and NE-AMRI protocols involved in this study did not participate in the multidisciplinary team, as well as clinical diagnosis and treatment work. The imaging analysis of our study was conducted separately and was independent of the clinical procedures. Therefore, the biopsy or surgery was performed according to the final consensus results of the multidisciplinary team and not affected by the evaluation results of CE-CT and NE-AMRI protocols involved in this study. Those with pathological results of the biopsy or operative specimens would be used as final reference standards. Furthermore, the operative recording, including the recording about the number and location of lesions discovered during surgery, as well as intraoperative contrast-enhanced ultrasound recording, could also be helpful to judge. All pathologic reports in our center were determined in consensus by two pathologists who underwent special training on colorectal carcinoma and liver metastasis pathology. The results consisted of the presence and number of CRLMs, the location and size of each lesion, and the date of first detection. Patients with lesions considered positive on CE-CT or NE-AMRI and fulfilling the diagnostic criteria of liver metastases on reference standards were regarded as true positives. Patients without lesions on CE-CT or NE-AMRI and reference standards were regarded as true negatives. Patients with CRLM diagnosed by reference standards but not classified as positive by CE-CT or NE-AMRI were regarded as false negatives. Lesions only positive on CE-CT or NE-AMRI were considered false positives.
Statistical analysis
SPSS Statistics (version 26) and Medcalc software (version 20) were utilized for all statistical analyses. A P-value less than 0.05 was considered significant. Interobserver agreement of diagnosed lesions between two radiologists was determined using kappa statistics. The detection rates of CRLM lesions (all lesions and lesions ≤ 10 mm) between two protocols were compared by the Chi-squared test or Fisher’s exact test. A subgroup analysis was performed in liver metastases in which size ≤ 10 mm. The per-patient sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated. The patient-based diagnostic performance of CE-CT and NE-AMRI in each round was evaluated by receiver operating characteristic (ROC) analyses, and the areas under the curves (AUCs) were calculated and compared using Delong’s test. Multivariable generalized linear mixed models were used to determine whether the diagnostic accuracy of CE-CT and NE-AMRI changed with patient age, gender, BMI, or fatty liver disease. Sankey diagrams were drawn comparing the flow and proportion of CE-CT and NE-AMRI protocols in diagnosing all metastatic lesions in three rounds. This intuitive visualization method allows observers to quickly capture the correlations between variables in different groups.
Results
Characteristics of primary enrolled patients and CRC tumors
As shown in Table 1, 587 patients were initially enrolled in the study of surveillance. There were 240 women (40.9%) and 347 men (59.1%) with a median age of 61 years (range, 32–83 years) and a mean BMI of 22.3 (SD, 6.1), wherein 183 patients (31.2%) suffered the fatty liver disease. The tumors of 151 patients (25.7%) were located on the right-sided colon, 297 (50.6%) were located on the left-sided colon, and 139 (23.7%) were located on the rectum. The serum CEA and CA199 levels were also recorded. The tumors of 84 patients (14.3%) presented with the T1/2 stage, and 503 (85.7%) presented with the T3/4 stage. In addition, the tumors of 261 patients (44.5%) presented with the N0 stage, and 326 patients (55.5%) presented with the N1/2 stage. 283 patients (48.2%) had KRAS mutation, and 21 patients (3.6%) had BRAF mutation.
Table 1.
Characteristics of primary enrolled patients and CRC tumors
| Characteristics | Value |
|---|---|
| Number of patients | 587 |
| Age (y)a | 61 (31-78) |
| Sex female/male | 240 (40.9)/347 (59.1) |
| BMIb | 22.3 (6.1) |
| Fatty liver disease Y/N | 183 (31.2)/404 (68.8) |
| CEA ≤200/>200 ng/mL | 419 (71.4)/168 (28.6) |
| CA199 ≤200/>200 U/mL | 394 (67.1)/193 (32.9) |
| Primary tumor location | |
| Right-sided/left-sided/rectum | 151 (25.7)/297 (50.6)/139 (23.7) |
| Primary tumor (T) stage | |
| T1/2/T3/4 | 84 (14.3)/503 (85.7) |
| Primary nodal (N) stage | |
| N0/N1/2 | 261 (44.5)/326 (55.5) |
| KRAS mutated Y/N | 283 (48.2)/304 (51.8) |
| BRAF mutated Y/N | 21 (3.6)/566 (96.4) |
Y = yes, N = no. Unless otherwise specified, data are numbers of patients, with percentage in parentheses.
Data are medians, with ranges in parentheses.
Data are means, with standard deviations in parentheses. Calculated as weight in kilograms divided by height in meters squared.
Characteristics of diagnosed CRLM in each round
Table 2 summarizes characteristics of diagnosed liver metastases in each round of surveillance. Taking round 1 as an example, 50 of 559 patients were diagnosed with 148 CRLM lesions. There were 14 women (28.0%) and 36 men (72.0%) with a median age of 57 years (range, 33–78 years) and a mean BMI of 22.6 (SD, 4.5), wherein 15 patients (30.0%) suffered the fatty liver disease. 21 of them (42.0%) had one liver metastasis, 24 (48.0%) had two to five metastases, and 5 (10.0%) had six to ten metastases. The lesion distribution of 27 patients (54.0%) was unilobar, and 23 (46.0%) was bilobar. The median time interval of CE-CT and NE-AMRI examinations was 2 days (range, 0–14 days). During this round of surveillance, 10 patients (20.0%) simultaneously experienced extrahepatic metastases. Regarding of the tumor size, 42 lesions (29.4%) were ≤10 mm, 57 (38.5%) were 10–20 mm, and 49 (33.1%) were ≥20 mm.
Table 2.
Characteristics of diagnosed CRLM in each round of surveillance
| Parameters | Round 1 | Round 2 | Round 3 |
|---|---|---|---|
| Pre-patient | |||
| Total number of diagnosed patients | 50 | 42 | 27 |
| Age (y) a | 57 (33-78) | 60.5 (30-77) | 54 (33-75) |
| Sex, female/male | 14 (28.0)/36 (72.0) | 16 (38.1)/26 (61.9) | 13 (48.1)/14 (51.9) |
| BMI b | 22.6 (4.5) | 21.9 (5.2) | 21.4 (3.8) |
| Fatty liver disease Y/N | 13 (26.0)/37 (74.0) | 12 (28.6)/30 (71.4) | 9 (33.3)/18 (74.1) |
| Number of CRLM | |||
| 1 | 21 (42.0) | 16 (38.1) | 9 (33.3) |
| >1, ≤5 | 24 (48.0) | 21 (50.0) | 14 (51.9) |
| >5, ≤10 | 5 (10.0) | 5 (11.9) | 4 (14.8) |
| Distribution | |||
| Unilobar | 27 (54.0) | 23 (54.8) | 16 (59.3) |
| Bilobar | 23 (46.0) | 19 (45.2) | 11 (40.7) |
| Days between NE-AMRI and CE-CT a | 2 (0-14) | 2 (0-8) | 3 (0-12) |
| Extrahepatic metastases Y/N | 10 (20.0)/40 (80.0) | 7 (16.7)/35 (83.3) | 6 (22.2)/21 (77.8) |
| Pre-lesion | |||
| Total number of lesions | 148 | 132 | 89 |
| Proof | |||
| Histopathology | 115 (77.7) | 95 (72.0) | 70 (78.7) |
| CE-MRI | 33 (22.3) | 37 (28.0) | 19 (21.3) |
| Tumor size | |||
| <10 mm | 42 (29.4) | 45 (34.1) | 29 (32.6) |
| 10–20 mm | 57 (38.5) | 52 (39.4) | 38 (42.7) |
| >20 mm | 49 (33.1) | 35 (26.5) | 22 (24.7) |
Y = yes, N = no. Unless otherwise specified, data are numbers of patients, with percentage in parentheses.
Data are medians, with ranges in parentheses.
Data are means, with standard deviations in parentheses. Calculated as weight in kilograms divided by height in meters squared.
Lesion-based detection rates
On average, in round 1, 80.4% metastases were detected with CE-CT, and 86.8% were detected with NE-AMRI (P < 0.001). In a subgroup analysis, for lesions ≤10 mm, 57.2% metastases were detected with CE-CT and 76.2% were detected with NE-AMRI (P = 0.017 for reviewer 1, P = 0.001 for reviewer 2). One of the lesions detected by NE-AMRI during this round is exemplified in Figure 2, and the lesion was not clearly displayed on CE-CT images but had typical signals on T2WI, DWI, and ADC. The results of rounds 2 and 3 are shown in Table 3. Furthermore, the images of one patient with fatty liver disease are listed in Figure 3, and the liver metastases displayed typical signals on NE-AMRI but were almost not displayed on CE-CT images. Regarding the interobserver agreements, the kappa value ranged from 0.796 to 0.929, confirming that there was great interobserver agreement for the evaluation between the two radiologists. In addition, the results of Sankey diagrams showed the data flow and proportional relationships between two protocols (Fig. 4). In the diagram, the width of each variable represents the quantity of data, while the color represents the type of data. Observers can understand how data flow between different groups by the direction and color of the arrows. Taking the results of reviewer 1 for example, the diagrams presented that 325 metastases were detected by NE-AMRI, but only 301 were detected by CE-CT. In addition, we can see intuitively that 41 metastases were misdiagnosed as hepatic cysts by the CE-CT protocol, but more than half of these 41 lesions were correctly identified by the NE-AMRI protocol. And 44 metastases were misclassified by the NE-AMRI protocol, of which only a few lesions were correctly identified by the CE-CT protocol.
Figure 2.
A 58-year-old female was confirmed CRLM in the first round of surveillance. The lesion was unclearly displayed on (A) axial plain and (B) arterial phase CT image. (C) The axial portal venous phase CT image showed an extremely small low-density point in the junction of left and right lobes of the liver (white arrow). The lesion showed a mild-moderate high-signal intensity on (D) T2WI, a higher signal on (E) DWI (b = 800), and a lower signal than the adjacent liver on (F) the corresponding ADC map (white arrow).
Table 3.
Lesion-based detection rates of two image sets by two reviewers in each round
| CE-CT | NE-AMRI | P value | ||
|---|---|---|---|---|
| Round 1 | ||||
| All lesions | Reviewer 1 | 81.7 (121/148) | 88.5 (131/148) | < 0.001 |
| Reviewer 2 | 79.1 (117/148) | 85.1 (126/148) | < 0.001 | |
| aк values | 0.914 (0.832–0.996) | 0.853 (0.728–0.978) | ||
| ≤10 mm lesions | Reviewer 1 | 61.9 (26/42) | 78.6 (33/42) | 0.017 |
| Reviewer 2 | 52.4 (22/42) | 73.8 (31/42) | 0.001 | |
| aк values | 0.807 (0.631–0.983) | 0.869 (0.693–1.000) | ||
| Round 2 | ||||
| All lesions | Reviewer 1 | 80.3 (107/132) | 87.9 (116/132) | < 0.001 |
| Reviewer 2 | 78.8 (104/132) | 85.6 (113/132) | < 0.001 | |
| aк values | 0.929 (0.851–1.000) | 0.901 (0.791-1.000) | ||
| ≤10 mm lesions | Reviewer 1 | 62.2 (28/45) | 77.8 (35/45) | < 0.001 |
| Reviewer 2 | 57.8 (26/45) | 71.1 (32/45) | 0.002 | |
| aк values | 0.908 (0.783–1.000) | 0.826 (0.638–1.000) | ||
| Round 3 | ||||
| All lesions | Reviewer 1 | 82.0 (73/89) | 87.6 (78/89) | 0.003 |
| Reviewer 2 | 75.3 (67/89) | 84.3 (75/89) | < 0.001 | |
| aк values | 0.830 (0.687–0.973) | 0.861 (0.708–1.000) | ||
| ≤10 mm lesions | Reviewer 1 | 65.6 (19/29) | 75.9 (22/29) | 0.003 |
| Reviewer 2 | 55.2 (16/29) | 72.4 (21/29) | 0.010 | |
| aк values | 0.786 (0.563–1.000) | 0.910 (0.738–1.000) | ||
Data are percentages, with fractions in parenthesis.
Data are к values, with 95% CI in parenthesis.
Figure 3.
A 59-year-old male with fatty liver disease was confirmed CRLM in the second round of surveillance. There was no lesion found on (A) the axial plain, (B) arterial phase, and (C) portal venous phase CT image. The lesion showed a mild-moderate high-signal intensity on (D) T2WI, a higher signal on (E) DWI (b = 800), and a lower signal than the adjacent liver on (F) the corresponding ADC map (white arrow).
Figure 4.
Sankey diagrams of two reviewers’ diagnoses. The specific diagnoses of all metastatic lesions in three rounds by the CE-CT protocol and NE-AMRI protocol were compared.
Patient-based diagnostic performance
The patient-based diagnostic performance of two image sets for liver metastases is summarized in Table 4. Overall, during three rounds of surveillance, the sensitivity, specificity, PPV, NPV, and accuracy of NE-AMRI were all higher than those of CE-CT. Typical images of two protocols of a patient without CRLM in the first round of surveillance were presented (Figure S1, http://links.lww.com/JS9/D799). The lesion was diagnosed as the liver metastases in CE-CT because of suspicious irregular peripheral enhancement in portal venous phases but had an obviously high-signal intensity on T2WI, a mild-moderate high signal on DWI, and a higher signal than the adjacent liver on the corresponding ADC map of NE-AMRI, which manifested as the hepatic cavernous hemangioma. The ROC curves of two image sets in three rounds are presented in Figure 5. The AUCs of NE-AMRI for three rounds were all higher than those of CE-CT (P = 0.015, 0.045, 0.009, respectively). In addition, we used multivariable generalized linear mixed models (binary logit) to explore the relationship between the diagnostic accuracy of CE-CT and NE-AMRI with patient age, gender, BMI, and fatty liver disease, respectively (Table 5). The results showed that the diagnostic accuracy of NE-AMRI did not change with any variables, but patient BMI and fatty liver disease had a significant impact on the diagnostic accuracy of CE-CT.
Table 4.
Patient-based diagnostic performance of two image sets in each round
| CE-CT | NE-AMRI | |
|---|---|---|
| Round 1 | ||
| Sensitivity (%) | 72.0 [57.5–83.8] | 86.0 [73.3–94.2] |
| Specificity (%) | 85.8 [82.7–88.6] | 90.6 [88.0–92.9] |
| PPV | 30.3 [25.0–36.1] | 43.9 [37.3–50.7] |
| NPV | 97.3 [95.8–98.2] | 98.7 [97.4–99.3] |
| Accuracy | 84.6 [81.9–87.6] | 90.2 [87.9–92.6] |
| Round 2 | ||
| Sensitivity (%) | 76.2 [60.5–87.9] | 88.1 [74.4–96.0] |
| Specificity (%) | 83.6 [79.9–86.8] | 87.5 [84.5–90.7] |
| PPV | 29.4 [24.2–35.1] | 39.4 [33.2–45.9] |
| NPV | 97.5 [95.8–98.5] | 98.8 [97.3–99.5] |
| Accuracy | 82.7 [79.3–85.9] | 87.4 [84.6–90.4] |
| Round 3 | ||
| Sensitivity (%) | 70.4 [49.8–86.2] | 88.9 [70.8–97.6] |
| Specificity (%) | 84.2 [79.9–88.0] | 90.5 [86.8–93.4] |
| PPV | 26.4 [20.2–33.7] | 42.9 [34.5–51.7] |
| NPV | 97.3 [95.2–98.4] | 99.0 [97.2–99.7] |
| Accuracy | 83.2 [79.3–87.1] | 90.3 [87.3–93.4] |
Data are percentages, with 95% CI in parenthesis.
Figure 5.
Receiver operating characteristic (ROC) curves of each image set in each round. The patient-based diagnostic performance of the NE-AMRI protocol was all significantly better than that of the CE-CT protocol.
Table 5.
Multivariable generalized linear mixed models (binary logit) of diagnostic accuracy with characteristics of patients
| Round 1 | Round 2 | Round 3 | ||||
|---|---|---|---|---|---|---|
| OR (95% CI) b | P value | OR (95% CI) b | P value | OR (95% CI) b | P value | |
| CE-CT | ||||||
| Age, y | 0.99 (0.97–1.01) | 0.211 | 1.05 (0.99–1.12) | 0.831 | 1.04 (1.02–1.06) | 0.800 |
| Female vs male | 1.37 (0.84–2.23) | 0.213 | 1.63 (0.92–2.90) | 0.094 | 1.10 (0.53–2.31) | 0.799 |
| BMIa of patient | 0.79 (0.36–1.81) | <0.001 | 0.55 (0.32–0.94) | 0.027 | 0.83 (0.81–0.86) | 0.002 |
| Fatty liver disease | 0.72 (0.41–1.35) | 0.005 | 0.47 (0.25–0.90) | 0.020 | 0.77 (0.35–1.79) | 0.009 |
| NE-AMRI | ||||||
| Age, y | 0.98 (0.95–1.01) | 0.522 | 1.01 (0.98–1.04) | 0.755 | 0.94 (0.51–1.57) | 0.656 |
| Female vs male | 1.06 (0.94–1.17) | 0.325 | 1.72 (0.93–3.19) | 0.084 | 1.21 (0.46–2.42) | 0.894 |
| BMIa of patient | 0.75 (0.22–2.11) | 0.087 | 0.64 (0.29–1.27) | 0.175 | 0.91 (0.87–0.95) | 0.063 |
| Fatty liver disease | 0.67 (0.41–1.07) | 0.054 | 0.56 (0.33–0.94) | 0.103 | 0.88 (0.75–1.02) | 0.080 |
Calculated as weight in kilograms divided by height in meters squared.
OR = odds ratio. ORs more than 1 indicate higher accuracy and ORs less than 1 indicate lower accuracy.
Discussion
On average, the observers misclassified 74 metastases in the CE-CT imaging set of three rounds together as benign lesions in our study. Among them, there are about 65% lesions less than 10 mm. Taking the lesion in Figure 2 as an example, it was neglected in CE-CT images due to its small scale but obviously displayed in NE-AMR images. A portion of benign lesions are similarly considered to be metastases in the study, and most of them were considered as hepatic cysts, hemangioma, or non-specific inflammation. Therefore, although CE-CT has been the common radiological tool for screening the occurrence of distant metastases of CRC after surgery[19,20], the sensitivity and NPV of it seemed to be relatively low. The overwhelming majority of the reason is owing to the partial volume effect of too small lesions in CT images[21]. Several research studies have also confirmed our viewpoints. Mueller et al. indicated that small lesions that were considered indeterminate on CE-CT were found to be malignant in approximately 22.4% of cases[22]. Furthermore, frequent CT scans increase radiation exposure. The cumulative radiation caused by ionizers and iodinated contrast agents could increase the risk of hematologic and other solid organ cancers[23,24], which limited CE-CT as a periodic surveillance tool for CRC patients. Therefore, methods to reduce radiation doses while maintaining diagnostic accuracy are important in this population.
Several clinical guidelines recommend the use of CE-MRI for improving capability for detection of small lesions in comparison with CE-CT imaging[25,26]. Burak et al. indicated that liver CE-MRI should be considered in all patients scheduled for local treatment for colorectal liver metastases on the basis of CE-CT imaging[27,28]. However, not all clinicians agreed with this strategy because the application of CE-MRI varies by institution. Some studies demonstrated that MRI had superiority for the detection of CRLM because of the inclusion of the T2WI and DWI sequence[29-31]. Therefore, we consider that non-contrast abbreviated MRI, only including T1WI, T2WI, DWI, and ADC maps, also has superiority for the detection of CRLM. In our study, the highest detection rate of lesions ≤10 mm by the NE-AMRI protocol could reach about 78.6% during three rounds of surveillance but only 65.6% by the CE-CT protocol. Most early lesions are quite small, and these small lesions are relatively easy to resect for surgeons. Therefore, a higher detection rate can avoid delaying opportunities of treatment, leading to disease progression. Of course, there will also be some over-diagnosis on these small lesions, particularly regarding false positives and unnecessary interventions. A comparative study confirmed that diffusion-weighted imaging had already showed significantly higher PPV (60%) than CE-CT (33%) in the detection of small liver metastases (≤10 mm diameter), confirming less false positives of DWI[30]. In addition, some studies indicated that CE-CT in combination with NE-MRI provided better diagnostic performance than CE-CT alone for detection of CRLM and similar to CE-MRI[7]. But it seems inappropriate to apply two types of examinations simultaneously as common radiological tools for screening the occurrence of liver metastases of CRC after surgery. From the Sankey diagrams, we found most of the misdiagnosed metastases were classified to hepatic cysts and hemangioma by the CE-CT protocol, which could be more easily recognizable by T2WI and DWI of the NE-AMRI protocol. Consequently, we further compared the diagnostic performance of CE-CT and NE-AMRI during three rounds of surveillance of CRLM. In our study, the performance of NE-AMRI were all proved to be higher than that of CE-CT, and the AUCs of alone NE-AMRI have already reached average 0.887 in three rounds.
Considering the lengthy examination time, high cost, and side effects associated with contrast agents, of CE-MRI, our strategy seems acceptable. If we use the cost of CE-MRI as a reference, in that way, the cost of NE-AMRI is about 60% of it, and the cost of CE-CT is approximately 55% of it. So the cost of NE-AMRI is similar to that of CE-CT, but NE-AMRI has obviously better diagnostic performance than CE-CT. Therefore, from the perspective of cost-effectiveness, it can be preliminarily seen that the NE-AMRI protocol may be preferable in clinical settings. In the clinical follow-up after the treatment of CRC, newly developed liver lesions are more likely to be metastases rather than benign lesions, such as hemangioma and cysts. Given that the NPV of NE-AMRI for detecting CRLM by patient-based analysis is up to 98.8%, active implementation of NE-AMRI is recommendable because characterization of benign lesions would be no more challenging. Furthermore, in our study, we found that the diagnostic accuracy of CE-CT could change with patient BMI and fatty liver disease, but NE-AMRI not. Patients with high BMI are more likely to develop fatty liver disease. With the changes in the lifestyle and diet, the incidence of non-alcoholic fatty liver disease (NAFLD) in adults is increasing, and the worldwide prevalence of NAFLD is 32.4%[32]. Besides, most patients with CRC would develop liver steatosis after chemotherapy and radical surgery, which reduced the sensitivity of CE-CT for small lesions[33,34]. In the third round of our study, the proportion of those with fatty liver disease had reached 33.3% among all diagnosed patients. Therefore, for these patients, reduced density of the liver parenchymal may affect the detection of small liver metastases by CE-CT, and NE-AMRI may have more advantage in the diagnosis of CRLM.
Our study has the following main limitations. First, the NE-AMRI protocol has limitations in assessing the anatomical topography for surgical planning[16]. Regarding the intrinsic limitation of NE-AMRI, we addressed it by performing corresponding other preoperative examinations once diagnosed in clinical practice. We agree that the NE-AMRI protocol is not superior to CE-MRI when highly exact anatomical information, such as the relationship between tumors and surrounding blood vessels, is needed. However, our study aimed to recommend an appropriate method for periodic surveillance of liver metastases in colorectal cancer patients after surgery, which emphasized the detection of lesions. Considering the high cost, lengthy examination time, side effects associated with contrast agents of CE-MRI, the NE-AMRI protocol seems acceptable in this long-term screening process; therefore, this intrinsic limitation of NE-AMRI has a relatively small impact on our study. Of course, in our study, once a patient was diagnosed with liver metastasis, doctors would develop personalized clinical strategies and perform corresponding other preoperative examinations based on the patient’s condition. Second, the relatively short time of follow-up was also a shortcoming of the study. We performed total 1.5 years of follow-up, mainly because the total count of population and the incidence of liver metastasis. Clinical data indicate that most colorectal liver metastases occur within 2 years after hepatectomy, and the incidence rate has gradually decreased with time[35]. The results observed in our study are similar to it. When reaching the third round, our patient cohort had been biased by the relatively small number of metastases, especially for lesions measuring ≤10 mm; the total count of metastases was only twenty-nine. Furthermore, in every round of our study, a small number of patients would be lost to follow-up due to the following two reasons: (a) withdrawn from the study or died without subsequent follow-up examinations and (b) unable to cooperate resulting in incomplete inspection or poor image quality. Most of these patients are older, with poor overall physical condition and compliance. Therefore, it may induce potential biases, resulting in a decrease in the representative of our sample. For these patients in clinical practice, more cautions should be exercised in the selection of the periodic surveillance protocol, such as training in MR scanning in advance, to avoid ineffective examinations and increasing the economic burden on patients. Next, this is a single-center study, and there is a lack of external validation. Our hospital has the high-volume colorectal cancer diagnosis and treatment center in terms of the annual number of surgeries performed on patients. Therefore, there was a comparatively wide range of populations with diverse conditions in our study. In addition, different treatment groups and four different scanning machines relatively established the reliability of NE-AMRI in different clinical settings. Our future studies would include multi-center validation and extend the follow-up duration to ensure generalizability and to address concerns about long-term clinical relevance.
Conclusion
This study demonstrated that NE-AMRI showed significantly better diagnostic performance than CE-CT as a surveillance tool for liver metastases in a cohort of stage II-III CRC patients within 18 months follow-up after surgery. NE-AMRI may be a promising periodic surveillance tool for CRC patients after surgery to increase the diagnostic accuracy of liver metastases, developing personalized clinical management and improving prognosis, simultaneously avoiding side effects associated with ionizing radiation and contrast agents.
Footnotes
Supplemental Digital Content is available for this article. Direct URL citations are provided in the HTML and PDF versions of this article on the journal’s website, www.lww.com/international-journal-of-surgery.
Jingjing Liu and Lechi Ye contributed equally to this work.
Published online 28 January 2025
Contributor Information
Jingjing Liu, Email: crystal-liu0123@163.com.
Gengyun Miao, Email: miaogengyun@163.com.
Shengxiang Rao, Email: shengxiang-med@126.com.
Mengsu Zeng, Email: mengsu-med@126.com.
Liheng Liu, Email: henryleozs@126.com.
Ethical approval
The study was approved by the institutional review board Zhongshan Hospital, Fudan University (No. Y2020-120).
Consent
Written informed consent was obtained from the patient for publication of this case report and accompanying images. A copy of the written consent is available for review by the Editor-in-Chief of this journal on request.
Sources of funding
This work was supported by the National Natural Science Foundation of China (Nos. 82371920 and 82402223), the China Postdoctoral Science Foundation (No. 2023M730722), the Shanghai Sailing Program (No. 23YF1441800), and the Outstanding Resident Clinical Postdoctoral Program of Zhongshan Hospital Affiliated to Fudan University (No. 2024ZYYS-016).
Author’s contribution
Conceptualization and data acquisition: all authors. Investigation, writing – original draft: J.J.L., L.C.Y., and L.H.L. Methodology: J.J.L., L.C.Y., M.S.Z., and L.H.L. Writing – review and editing: J.J.L. and L.H.L. Data analysis and interpretation: J.J.L., L.C.Y., G.Y.M., and L.H.L. Study administration: M.S.Z., and L.H.L. Final approval of publication: J.J.L. and L.H.L.
Conflicts of interest disclosure
The authors declare that they have no conflicts of interest.
Research registration unique identifying number (UIN)
No. ChiCTR2000034228.
Guarantor
Liheng Liu had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Provenance and peer review
Not commissioned, externally peer-reviewed.
Data availability statement
Data are available upon reasonable request.
Assistance with the study
None.
Presentation
None.
References
- [1].Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2021;71:209–49. [DOI] [PubMed] [Google Scholar]
- [2].Cervantes A, Adam R, Roselló S, et al. Metastatic colorectal cancer: ESMO clinical practice guideline for diagnosis, treatment and follow-up. Ann Oncol 2023;34:10–32. [DOI] [PubMed] [Google Scholar]
- [3].Giannis D, Sideris G, Kakos CD, Katsaros I, Ziogas IA. The role of liver transplantation for colorectal liver metastases: a systematic review and pooled analysis. Transplant Rev (Orlando) 2020;34:100570. [DOI] [PubMed] [Google Scholar]
- [4].Pitroda SP, Chmura SJ, Weichselbaum RR. Integration of radiotherapy and immunotherapy for treatment of oligometastases. Lancet Oncol 2019;20:e434–e442. [DOI] [PubMed] [Google Scholar]
- [5].Baca B, Beart RW,J, Etzioni DA. Surveillance after colorectal cancer resection: a systematic review. Dis Colon Rectum 2011;54:1036–48. [DOI] [PubMed] [Google Scholar]
- [6].Bregni G, Akin Telli T, Camera S, et al. Grey areas and evidence gaps in the management of rectal cancer as revealed by comparing recommendations from clinical guidelines. Cancer Treat Rev 2020;82:101930. [DOI] [PubMed] [Google Scholar]
- [7].Qiu QS, Zhu K, Wang JH, et al. Diagnostic performance of contrast enhanced CT alone or in combination with (non-)enhanced MRI for colorectal liver metastasis. Acad Radiol 2023;30:1856–65. [DOI] [PubMed] [Google Scholar]
- [8].Niekel MC, Bipat S, Stoker J. Diagnostic imaging of colorectal liver metastases with CT, MR imaging, FDG PET, and/or FDG PET/CT: a meta-analysis of prospective studies including patients who have not previously undergone treatment. Radiology 2010;257:674–84. [DOI] [PubMed] [Google Scholar]
- [9].Brenner DJ, Hall EJ. Computed tomography–an increasing source of radiation exposure. N Engl J Med 2007;357:2277–84. [DOI] [PubMed] [Google Scholar]
- [10].Mao Y, Chen B, Wang H, et al. Diagnostic performance of magnetic resonance imaging for colorectal liver metastasis: a systematic review and meta-analysis. Sci Rep 2020;10:1969. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Rogosnitzky M, Branch S. Gadolinium-based contrast agent toxicity: a review of known and proposed mechanisms. Biometals 2016;29:365–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Levine D, McDonald RJ, Kressel HY. Gadolinium retention after contrast-enhanced MRI. Jama 2018;320:1853–54. [DOI] [PubMed] [Google Scholar]
- [13].Mouawad M, Biernaski H, Brackstone M, et al. Reducing the dose of gadolinium-based contrast agents for DCE-MRI guided SBRT: the effects on inter and intra observer variability for preoperative target volume delineation in early stage breast cancer patients. Radiother Oncol 2019;131:60–65. [DOI] [PubMed] [Google Scholar]
- [14].Chan MV, Huo YR, Trieu N, et al. Noncontrast MRI for hepatocellular carcinoma detection: a systematic review and meta-analysis – a potential surveillance tool? Clin Gastroenterol Hepatol 2022;20:44–56.e2. [DOI] [PubMed] [Google Scholar]
- [15].Park SH, Kim B, Kim SY, et al. Characterizing computed tomography-detected arterial hyperenhancing-only lesions in patients at risk of hepatocellular carcinoma: can non-contrast magnetic resonance imaging be used for sequential imaging? Korean J Radiol 2020;21:280–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Hwang JA, Kim YK, Min JH, Song KD, Sohn I, Ahn HS. Non-contrast liver MRI as an alternative to gadoxetic acid-enhanced MRI for liver metastasis from colorectal cancer. Acta Radiol 2019;60:441–50. [DOI] [PubMed] [Google Scholar]
- [17].Mathew G, Agha R, Albrecht J, et al. STROCSS 2021: strengthening the reporting of cohort, cross-sectional and case-control studies in surgery. Int J Surg 2021;96:106165. [DOI] [PubMed] [Google Scholar]
- [18].Hahn L, Reeder SB, Muñoz Del Rio A, Pickhardt PJ. Longitudinal changes in liver fat content in asymptomatic adults: hepatic attenuation on unenhanced CT as an imaging biomarker for steatosis. AJR Am J Roentgenol 2015;205:1167–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Simpson AL, Peoples J, Creasy JM, et al. Preoperative CT and survival data for patients undergoing resection of colorectal liver metastases. Sci Data 2024;11:172. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Tsili AC, Alexiou G, Naka C, Argyropoulou MI. Imaging of colorectal cancer liver metastases using contrast-enhanced US, multidetector CT, MRI, and FDG PET/CT: a meta-analysis. Acta Radiol 2021;62:302–12. [DOI] [PubMed] [Google Scholar]
- [21].Sahani DV, Bajwa MA, Andrabi Y, Bajpai S, Cusack JC. Current status of imaging and emerging techniques to evaluate liver metastases from colorectal carcinoma. Ann Surg 2014;259:861–72. [DOI] [PubMed] [Google Scholar]
- [22].Mueller GC, Hussain HK, Carlos RC, Nghiem HV, Francis IR. Effectiveness of MR imaging in characterizing small hepatic lesions: routine versus expert interpretation. AJR Am J Roentgenol 2003;180:673–80. [DOI] [PubMed] [Google Scholar]
- [23].Rizzo S, Bellesi L, D’Ermo A, et al. Body CT examinations in oncologic patients: the impact of subspecialty radiology on radiation exposure in the clinical practice. A quality care study. Radiol Med 2024;129:429–38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Wang L, Li Q, Wang XM, et al. Enhanced radiation damage caused by iodinated contrast agents during CT examination. Eur J Radiol 2017;92:72–77. [DOI] [PubMed] [Google Scholar]
- [25].Lin YM. Enhancing clinical decision-making: the impact of liver contrast-enhanced MRI on local treatment plans for colorectal liver metastases. Radiol Imaging Cancer 2024;6:e239023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [26].Attiyeh MA, Malhotra GK, Li D, Manoukian SB, Motarjem PM, Singh G. Defining MRI superiority over CT for colorectal and neuroendocrine liver metastases. Cancers (Basel) 2023;15:5109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Görgec B, Hansen IS, Kemmerich G, et al. MRI in addition to CT in patients scheduled for local therapy of colorectal liver metastases (CAMINO): an international, multicentre, prospective, diagnostic accuracy trial. Lancet Oncol 2024;25:137–46. [DOI] [PubMed] [Google Scholar]
- [28].Zech CJ, Korpraphong P, Huppertz A, et al. Randomized multicentre trial of gadoxetic acid-enhanced MRI versus conventional MRI or CT in the staging of colorectal cancer liver metastases. Br J Surg 2014;101:613–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [29].Sivesgaard K, Larsen LP, Sørensen M, et al. Diagnostic accuracy of CE-CT, MRI and FDG PET/CT for detecting colorectal cancer liver metastases in patients considered eligible for hepatic resection and/or local ablation. Eur Radiol 2018;28:4735–47. [DOI] [PubMed] [Google Scholar]
- [30].Kim HJ, Lee SS, Byun JH, et al. Incremental value of liver MR imaging in patients with potentially curable colorectal hepatic metastasis detected at CT: a prospective comparison of diffusion-weighted imaging, gadoxetic acid-enhanced MR imaging, and a combination of both MR techniques. Radiology 2015;274:712–22. [DOI] [PubMed] [Google Scholar]
- [31].Colagrande S, Castellani A, Nardi C, Lorini C, Calistri L, Filippone A. The role of diffusion-weighted imaging in the detection of hepatic metastases from colorectal cancer: a comparison with unenhanced and Gd-EOB-DTPA enhanced MRI. Eur J Radiol 2016;85:1027–34. [DOI] [PubMed] [Google Scholar]
- [32].Riazi K, Azhari H, Charette JH, et al. The prevalence and incidence of NAFLD worldwide: a systematic review and meta-analysis. Lancet Gastroenterol Hepatol 2022;7:851–61. [DOI] [PubMed] [Google Scholar]
- [33].Holalkere NS, Sahani DV, Blake MA, Halpern EF, Hahn PF, Mueller PR. Characterization of small liver lesions: added role of MR after MDCT. J Comput Assist Tomogr 2006;30:591–96. [DOI] [PubMed] [Google Scholar]
- [34].Miyata T, Shinden Y, Motoyama S, et al. Non-alcoholic fatty liver disease may be a risk factor for liver metastasis after radical surgery for colorectal cancer: a retrospective study. J Gastrointest Cancer 2024;55:932–39. [DOI] [PubMed] [Google Scholar]
- [35].Galjart B, van der Stok EP, Rothbarth J, Grünhagen DJ, Verhoef C. Posttreatment surveillance in patients with prolonged disease-free survival after resection of colorectal liver metastasis. Ann Surg Oncol 2016;23:3999–4007. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Data are available upon reasonable request.