Liver Tumor Imaging Staging: A Multi-Institutional Study of a Preoperative Staging Tool for Hepatocellular Carcinoma

doi:10.21203/rs.3.rs-5013690/v1

Download PDF

Research Article

Liver Tumor Imaging Staging: A Multi-Institutional Study of a Preoperative Staging Tool for Hepatocellular Carcinoma

https://doi.org/10.21203/rs.3.rs-5013690/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background & Aims:

The current staging system has limitations in preoperatively assessing hepatocellular carcinoma (HCC) and in precise detailed treatment allocation. This study aims to propose a new Liver Tumor Imaging Staging (LTIS) method for HCC.

Methods

1295 patients who underwent CT or MRI and curative liver resection during January 2012 and October 2020 were retrospectively recruited from three independent institutions. LTIS was designed to discriminate low-grade (absence of microvascular invasion [MVI] and Edmondson-Steiner grade III/IV), intermediate (MVI + or Edmondson-Steiner grade III/IV but not both) and high-grade HCC (MVI + and Edmondson-Steiner grade III/IV) upon CT and MRI. Model was constructed in 578 derivation cohort (center 1) and validated in internal center 1 test cohort (n = 291), and external center 2 (n = 226) and center 3 (n = 200), respectively. Net clinical benefit of LTIS on recurrence-free survival (RFS) and overall survival (OS) was analyzed with a Cox proportional hazards model.

Results

In independent test, LTIS achieved agreement of 73.2% (281/384), 18.9% (100/528), and 69.2% (265/383) for determining low, intermediate, and high-grade HCCs with “ground truth” results. In the Cox analysis, LTIS was comparable to “ground truth” grade for predicting RFS (hazards ratio (HR), 1.30 vs ground truth grade, 1.36 and 1.56) and OS (HR, 1.76 vs ground truth grade, 2.00 to 3.03) of patients after surgery. In patients conventionally classified as having low-grade tumors (serum α-fetoprotein < 40 ng/mL, stage T1), 47.4% and 35.6% were reclassified as high-grade tumors upon LTIS restaging. The resulting LTIS subgroups showed a significant difference in RFS and OS at Kaplan-Meier analysis (Log-rank test, p < 0.001).

Conclusion

LTIS provides a potential noninvasive way to precisely stage HCC using CT and MRI.

Hepatocellular carcinoma

liver tumor imaging staging

disease prediction

clinical decision support

Hepatocellular carcinoma (HCC) ranks as the fifth most frequently diagnosed cancer and the third leading cause of cancer-related mortality, with particularly high incidence and mortality rates in the Asia-Pacific region [1]. Advancements in techniques and treatments for hepatocellular carcinoma (HCC) have enabled patients to attain improved therapeutic outcomes [2]. However, the recurrence rate for early HCC is up to 50% within 2 years of resection and 12.1% patients die within 5 years after surgery [3].

Current staging systems, including Barcelona Clinic Liver Cancer (BCLC), Hong Kong Liver Cancer, Cancer of the Liver Italian Program(CLIP) and American Joint Committee on Cancer (AJCC) systems, have evolved to effectively stratify risk and facilitate precise treatment allocation [1]. However, the AJCC staging system depends solely on postoperative histopathological findings, thereby limiting its utility in preoperative treatment planning, The BCLC stage system predominantly depends on tumor number and size, yet it does not account for critical prognostic factors such as microvascular invasion (MVI), which distinguishes pT2 from pT1b in the AJCC 8th edition, or Edmondson-Steiner (ES) grade. Nevertheless, some studies suggest that there is potential for enhancing both staging and prognostic accuracy [4]. Imaging examinations such as CT and MRI could be a promising tool for detection, staging, and risk assessment of HCC [5]. The specific imaging phenotypes including tumor size, capsule, margin appearance and peritumoral enhancement features are highly related to histopathological characteristics of HCC such as MVI or ES grade [5–7]. Predictive models based on clinically supported indicators have been developed, enabling clinicians to select the most appropriate treatment for HCC patients prior to therapy[7–10]. Although these studies have reported promising results, their predictive models must be validated on external patients cohorts with different characteristics from the original dataset before clinical adoption. Additionally, there is a wide variation in model performance as it depends on various factors or imaging assessments.

We hypothesize that it is clinically feasible to determine appropriate preoperative signatures that can represent MVI and ES grade, with the aim of emulating the diagnostic precision of histopathology in staging and grading HCC. Therefore, a Liver Tumor Imaging Staging (LTIS) approach based on CT and MRI was proposed and evaluated in staging and grading HCC in a multicenter study.

Study Population

This retrospective study involved three medical centers (center 1, the First Affiliated Hospital of *****University; center 2, *****University Medical College First Affiliated Hospital; center 3, Affiliated Cancer Hospital of *****University). Ethics committee approval was granted by the local institutional ethics review board (Grant No. 2019-SR-376), and the requirement of written informed consent was waived.

We queried the institution’s medical records to derive all pathologically confirmed cases of HCC between January 2012 and October 2020. A total of 1,295 consecutive patients with HCC who underwent curative liver resection were screened in this query. Among these, 869 patients from center 1, 226 patients from center 2 and 200 patients from center 3 who underwent preoperative CT and/or MRI examination before the surgery formed the primary cohort. All patients did not have the history of previous surgical or transcatheter arterial chemoembolization or other therapies. Of patients included, 941 patients underwent a contrast-enhanced CT and 354 patients underwent a Gd-DTPA or hepato-specific contrast agents Gd-EOB-DTPA (Magnevist & Primovist; Bayer HealthCare, Berlin, Germany) MRI examination within 1 month before the surgery. The detailed imaging parameters of CT and MRI was summarized in supplemental materials (suppl. text-1).

Data collection

Routine preoperative laboratory examinations included liver function tests, hepatitis B and C immunology, serum α-fetoprotein (AFP) level, Child-Pugh class, serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), serum bilirubin, serum albumin, platelet count, prothrombin time and HBV DNA load. The albumin-bilirubin (ALBI) score was computed by the formula: 0.085*(albumin) + 0.66*log (bilirubin), and was then stratified according to previously described cut-offs: ALBI grade 1 (≤ 2.60) and grade 2–3 (> 2.60) [11].

Histopathological examination was performed by six experienced hepatopathologists through serially examining multiple pathologic specimens. Histopathological parameters ordinarily included the Edmondson-Steiner grade (26), pathological stage using 8th edition of AJCC TNM staging system for liver cancer. Tumor size, number, surgical margin, and MVI of the resected tumor were also reported. HCCs were classified into three pathological grades (PG): PG 1 (absence of MVI and ES grade I/II), PG 2 (MVI + or ES grade III/IV), and PG 3 (MVI + and ES grade III/IV).

After curative surgery, all patients were followed up according to institutional practice including laboratory examinations such as serum AFP 3-monthly and CT/MRI every 3 to 6 months. Recurrence-free survival (RFS) was defined as the time from date of surgery to the time of recurrence. Recurrence was defined as a presence of tumor dissemination or development of de novo tumors in the cirrhotic liver after curative resection. Overall survival (OS) was computed from the date of surgery to date of death or censored at the date of last follow-up. Patients were censored in case of emigration, or on 31th Oct. 2020, whichever came first.

Image Analysis

A total of 578 patients randomly selected from the derivation cohort (center 1) were used for the development of LTIS system, and the remaining 291 patients were used for internal test. All images in derivation data were interpreted by two abdominal radiologists and a board-certified radiologist who has read more than 3,000 liver imaging studies. The interpretations between juniors and the senior were allocated to assess inter-reader agreement. The senior’s interpretations were used for model development. The senior repeated the review twice with 4-week interval to guarantee reproducibility.

All readers independently reviewed the images and were blinded to any information regarding clinical history, ultrasound findings or final diagnosis. The following imaging features were recorded: 1) tumor size: the maximum length measured on portal-venous phase image (< 3 cm, 3–5 cm, > 5 cm) ; 2) tumor margin (well-defined vs ill-defined); 3) tumor capsule appearance (complete vs incomplete); 4) intratumoral necrosis (absent vs present); 5) wash-in/washout feature (absent vs present), according to the Liver Imaging Reporting and Data System for nodules > 1 cm (non-rim arterial phase hyperenhancement and washout of contrast media in the portal-venous/delayed phases) [12]; 6) focal extra-nodular extension (absent vs present); 7) arterial peritumoral enhancement (APTE) (absent vs present); 8) star nodule (absent vs present), the satellite nodules are these lesions with size ≤ 2 cm and located ≤ 2 cm from the main tumor [13]; 9) a two-trait predictor of venous invasion (TTPVI) score (absent vs present), which was based on independent estimations from internal arteries and hypodense halos [5, 6].

Model Development and Validation

The clinical adoption of a tool required for easy-to-use by doctors should follow a continual trust-building approach from preclinical development to real-world clinical deployment. Here, we divided the interpretable LTIS tool into two steps. First, a multinomial logistic regression analysis was used to select the imaging features predictive of HCC aggressiveness and assess interactions. Second, to identify an easy-to-use diagnostic scheme, the results of the regression model are presented as a finally five-point LTIS score, which is intuitive and easier for the interpretation of HCC with CT and MRI (see a flowchart in Suppl Fig. 1).

Next, the clinical applicability of LTIS was independently tested in primary derivation cohort (n = 869) and two external validation cohorts from center 2 (n = 226) and center 3 (n = 200) with six radiologists. All the readers were not involved in the imaging analysis procedure for model development above-mentioned.

Net Clinical Benefit of LTIS and Postsurgical Prognosis

We evaluated a Cox’s proportional hazard model using LTIS, clinic-imaging features including age, sex, preoperative laboratory indicators, and AJCC stage to assess the incremental aspect of LTIS for predicting RFS and OS of HCC patients after curative resection. Additionally, in AJCC stage 8th, MVI was considered to discriminate between T1b and T2. Specially, we assessed the net clinical benefit of restaging by LTIS in patients with tumor size < 3cm, low or absent AFP (< 40 ng/mL), or AJCC stage T1, which are conventionally unlikely to be staged as high aggressive tumors.

Statistical Analysis

To assess interobserver agreement in LTIS assessment, Cronbach's alpha statistics were determined for LTIS grade reviewed by three independent readers. The performance of LTIS for pathological grade was analyzed using a receiver operating characteristic (ROC) curve. To allow comparisons and to determine the net benefit derived from the clinical usefulness of the new LTIS, we relied on a decision curve analysis as described by Vickers et al [14], as well as with a decrease in accuracy with a random forest analysis to plot individual importance of LTIS score versus clinically available predictors for pathological grade of HCC. A Kaplan-Meier’s plot with log-rank tests was performed to determine the RFS and OS of patients using LTIS score. Stepwise univariable and multivariable Cox proportional hazards analysis were used to determine the independent predictors of disease-specific recurrence and mortality. Statistical analysis was performed using the R software (version 3.4.4, R Project for Statistical Computing, http://www.r-project.org). Two-sided p values < 0.05 were considered statistically significant.

Baseline characteristics

Table 1 shows detailed baseline characcteristics of all patients. Of all patients included, PG 1, PG 2, and PG 3 of HCC were diagnosed in explanted tissue of 324/869 (37.3%), 327/869 (37.6%), 218/869 (25.1%) patients, respectively, in center 1; 17/226 (7.5%), 127/226 (56.2%), 82/226 (36.3%), respectively, in center 2; and 43/200 (21.5%), 74/200 (37.0%), 83/200 (41.5%), respectively, in center 3.

Table 1

The clinical and histologic characteristics of patients from three medical centers.
Characteristics	Cohorts
Characteristics	Center 1 (n = 869)	Center 2 (n = 226)	Center 3 (n = 200)
Imaging, n (%)
CT	686 (78.9)	189 (83.6)	66 (33.0)
MRI	183 (21.1)	37 (16.4)	134 (67.0)
Age, n (%)
< 55yrs	426 (49.0)	112 (49.6)	124 (62.0)
≥ 55yrs	443 (51.0)	114 (50.4)	76 (38.0)
Sex, n (%)
Male	730 (84.0)	177 (78.3)	173 (86.5)
Female	139 (16.0)	49 (21.7)	27 (13.5)
HBV/HCV, n (%)
Absent	121 (13.9)	64 (28.3)	12 (6.0)
Present	748 (86.1)	162 (71.7)	188 (94.0)
Child-Pugh
A	772 (89.0)	184 (81.4)	126 (96.0)
B/C	96 (11.0)	42 (19.6)	8 (4.0)
AFP, n (%)
< 400 ng/mL	568 (65.4)	159 (70.4)	109 (54.5)
≥ 400 ng/mL	301 (34.6)	67 (29.6)	91 (45.5)
ALB (g/L), n (%)
< 40 g/L	384 (44.2)	156 (69.0)	80 (40.0)
≥ 40 g/L	485 (55.8)	70 (31.0)	120 (60.0)
TB (umol/L), n (%)
< 17.1 umol/L	533 (61.3)	93 (41.2)	137 (68.5)
≥ 17.1 umol/L	336 (38.7)	133 (58.8)	63 (31.5)
ALBI, n (%)
low: ≤ -2.60	379 (43.6)	157 (69.5)	81 (40.5)
high: > -2.60	490 (56.4)	69 (30.5)	119 (59.5)
PT (s), n (%)
< 13 s	442 (50.9)	30 (13.3)	88 (44.0)
≥ 13 s	427 (49.1)	196 (86.7)	112 (56.0)
Tumor size, n (%)
< 5 cm	441 (50.7)	81 (35.8)	35 (17.5)
≥ 5 cm	428 (49.3)	145 (64.2)	165 (82.5)
pAJCC stage
I †	298 (34.3)	-	-
Ia	51 (5.9)	25 (11.1)	17 (8.5)
Ib	117 (13.5)	37 (16.4)	42 (21.0)
II	179 (20.6)	67 (29.6)	31 (15.5)
III †	58 (6.7)	97 (42.9)	110 (55.0)
IIIa	83 (9.6)	-	-
IIIb	78 (8.9)	-	-
IV	5 (0.5)	-	-
Surgical no. of tumors, n (%)
Solid	721 (83.0)	205 (90.7)	171 (85.5)
2	71 (8.2)	15 (6.6)	14 (7.0)
≥ 3	77 (8.8)	6 (2.7)	15 (7.5)
Edmondson-Steiner, n (%)
Ⅰ/Ⅰ-Ⅱ	395 (45.5)	100 (44.2)	92 (46.0)
Ⅱ-Ⅲ/Ⅲ	474 (54.5)	126 (55.8)	108 (54.0)
MVI, n (%)
Absent	580 (66.7)	61 (27.0)	68 (34.0)
Present	256 (29.5)	165 (73.0)	132 (66.0)
Pathological Grade, n (%)
PG 1 (MVI- & ES I/II)	324 (37.3)	17 (7.5)	43 (21.5)
PG 2 (MVI + or ES III/IV)	327 (37.6)	127 (56.2)	74 (37.0)
PG 3 (MVI+ & ES III/IV)	218 (25.1)	82 (36.3)	83 (41.5)
Note. -Unless indicated otherwise, data are number of tumors, with percentages in parentheses. †, Patients were assigned with an overall stag I or III, while the subtype Ia, Ib, IIIa or IIIb was not evaluated at the pathology. Albumin-bilirubin (ALBI) score = 0.085 * albumin (g/L) + 0.66 * log (bilirubin umol/l). HBV = hepatitis B virus; HCV = hepatitis C virus; AFP = serum α-fetoprotein; ALB = albumin; TB = total bilirubin; PT = prothrombin time; MVI = microvascular invasion.

Radiomic Feature Selection

The presence of APTE (odds ratio [LORs], 1.18; 95% CIs, 1.01 to 1.27), presence of star node (LORs, 0.90; 95% CIs, 0.78 to 1.13), ill-defined margin (LORs, 0.81; 95% CIs, 0.63 to 0.98), presence of TTPVI (LORs, 0.71; 95% CIs, 0.58 to 0.84), tumor size > 5cm (LORs, 0.63; 95% CIs, 0.53 to 0.81) were positively associated with PGs; while completed capsule (LORs, -1.84; 95% CIs, -2.07 to -1.67), well-defined margin (LORs, -1.52; 95% CIs, -1.81 to -1.34), tumor size < 3 cm (LORs, -1.41; 95% CIs, -1.69 to -1.22), and absence of TTPVI (LORs, -0.91; 95% CIs, -1.16 to -0.86) were negatively associated with PGs. The original multinomial regression model resulted in an AUC of 0.78 (95% CIs, 0.70–0.81), 0.64 (95% CIs, 0.58–0.72) and 0.87 (95% CIs, 0.84–0.91) for PG 1, PG 2, and PG 3, respectively, on training data (Fig. 1a). On internal test, the model resulted in an AUC of 0.79 (95% CIs, 0.71–0.83), 0.60 (95% CIs, 0.55–0.70) and 0.86 (95% CIs, 0.83–0.91) (Fig. 1b), respectively.

The classification model was transformed into a graphically illustrated diagnostic scheme. As shown in Fig. 2a and Fig. 2b, LTIS uses a five-point LTIS risk score to interpret probability of high pathological grade by combining independent imaging features determined at regression model. Inter-reader agreements of LTIS were assessed in randomly selected in 494 CT datasets and in 183 MRI datasets. LTIS shows good inter-reader agreements in both CT and MRI datasets, with a Cronbach's alpha coefficient of 0.86 (95% CIs, 0.83–0.89) and 0.85 (95% CI, 0.81–0.89), respectively.

Prediction Model Development and Validation

The incidence of “ground truth” PGs stratified by LTIS is depicted in Fig. 3a, with significant linear trends in incidence of PGs across LTIS scores (Cochran Armitage chi-square, p < 0.01). The distribution of pathological findings by LTIS categories in three sub cohorts is illustrated in Suppl Fig. 2. Using LTIS 3 as threshold (≥ 3) for diagnosing PG 2–3, sensitivity, specificity, positive predictive value, and negative predictive value is 73.3% (658/910), 73.2% (282/385), 86.5% (658/761), and 52.8% (282/534), respectively. Compared to conventional clinical factors such as age, sex, and serum biomarkers, the LTIS outperformed them with significantly higher impact index at the random forest analysis by measuring decrease in accuracy (Fig. 3b). Moreover, with the decision curve analysis, adding LTIS to these clinical variables produced obviously higher net clinical benefits for the evaluation of PG (Fig. 3c).

Net Clinical Benefit of LTIS Score on Postsurgical Prognosis

Among the patients entirely followed up, 482 patients presented with a recurrence after liver resection, with a median RFS of 25.3 months (95% CIs, 20.4–27.8 months). 255 patients died after the curative resection, with a median OS of 74.5 months (95% CIs, 54.4–81.4 months). A scatter plot between RFS and OS in subgroups of LTIS score 1 to 5 is summarized in Fig. 4. It shows that patients with higher LTIS score trended to has lower median RFS and lower median OS. In subgroup (n = 395) within 1-year recurrence and 2-year death after surgery, 51.1% (202/395) of patients had LTIS score 4 and score 5.

At the stepwise Cox’s proportional hazard model, the male (hazards ratio [HR], 1.45; 95% CIs, 1.11–1.90; p = 0.006), aspartate aminotransferase (AST) ≥ 40 U/L (HR, 1.32; 95% CIs, 1.09–1.59; p = 0.004), α-fetoprotein (AFP) ≥ 400 ng/mL(HR, 1.56; 95% CIs, 1.29–1.89; p < 0.001), total bilirubin (TB) ≥ 17.1 umol/L (HR, 0.75; 95% CIs, 0.62–0.91; p = 0.004), LTIS score 3–5 (HR, 1.30; 95% CIs, 1.03–1.65, p = 0.026), AJCC III/IV (HR, 1.56; 95% CIs, 1.21–2.01, p < 0.001) and PG 3 (HR, 1.36; 95% CIs, 1.02–1.82, p = 0.038) were identified as independent predictors of RFS. The resulting Cox model with independent predictors produces a C-index of 0.71 (95% CIs, 0.68–0.76) for predicting RFS (Suppl Fig. 3a). The AST ≥ 40 U/L (HR, 1.51; 95% CIs, 1.14–2.00; p = 0.004), AFP ≥ 400 ng/mL (HR, 1.38; 95% CIs, 1.06–1.80; p = 0.017), LTIS score 3–5 (HR, 1.76; 95% CIs, 1.23–2.50, p = 0.002), AJCC III/IV (HR, 3.03; 95% CIs, 2.14–4.30, p < 0.001) and PG 3 (HR, 2.00; 95% CIs, 1.32–3.00, p < 0.001) were identified as independent predictors of OS. The resulting Cox model produces a C-index of 0.76 (95% CIs, 0.72–0.79) for predicting OS (Suppl Fig. 3b). Specially, in patients with tumor size < 3 cm (n = 372), AFP < 40 ng/mL (n = 548), and AJCC stage T1 (n = 587), the rate of LTIS score 3–5 was 27.2%, 47.4% and 35.6% (Fig. 5a). HCC patients at AFP < 40 ng/mL and AJCC stage T1 restaged by LTIS represented with significantly (Log-rank test, p < 0.001) different RFS and OS at Kaplan-Meier survival curve analysis (Fig. 5b and Fig. 5c).

In present study, we proposed an intuitive LTIS tool to preoperatively assess pathological characteristics of HCC using CT and MRI. Our result demonstrate that LTIS assignment achieves high reproducibility and sufficient accuracy among multicenter high-variability datasets and across radiologists with vary ing levels of experiences. Importantly, this new staging model incorporates pathologically aggressive factors such as MVI and ES grade of HCC. Additionally, patients who are traditionally classified as having "low-aggressive" tumors benefit clinically from improved prediction of postsurgical outcomes when assessed using the LTIS.

MVI has recently been incorported into the AJCC staging scheme to discriminate between pT1b and pT2 tumors. Several single-center studies have demonstrated that, larger tumor size, unclear margin, incomplete capsule, extranodular extension, presence of APTE and TTPVI on CT and MRI are associated with a higher risk of MVI [5, 6, 9, 15]. Different criteria enrollment and assessment strategies can cause heterogeneity, resulting in sensitivity ranging from 11–100% and specificity from 5–97% in these studies. In this study, stepwise LTIS analysis revealed that tumor size, margin, capsule, APTE, TTPVI, and star nodules were associated with increased tumor aggressiveness. Notably, LTIS effectively predicts PG-3 diseases with the presence of MVI and ES grade III-IV with sufficient accuracy. This presurgical tool can provide an alternative for surgeons in selecting anatomic resection, wide resection margin and postoperative adjuvant transarterial chemoembolization for patients with high-aggressive HCCs [16]. In spite of presurgical diagnosis of MVI or aggressiveness has not yet been included in the latest versions of the guidelines for HCC management [3, 17, 18].

Early recurrence within 2 years after curative resection is currently the most robust clinical endpoint for assessing HCC [3, 4, 17]. It had correctly demonstrated that LTIS score was comparable to pAJCC stage and PG in assessing RFS and OS. Conventional staging systems often fail to classifiy patients with small (< 3 cm), solitary tumors and low or absent AFP as high-aggressive tumors, with limited research on their impact on postsurgical outcomes [19]. However, this study implied a potential risk of underestimating HCC aggressiveness, particularly in patients classified as “low-aggressive” by conventional methods. Approximately 27.2%, 47.4% and 35.6% of patients with tumor size < 3 cm, AFP < 40 ng/mL and AJCC stage T1, respectively, were identified as LTIS 3–5. Patients with AFP < 40 ng/mL and AJCC stage T1 restaged by LTIS showed significantly different postsurgical outcomes. Therefore, use of LTIS to restage patients with “low-aggressive” tumors might provide additional predictive value in postsurgical prognosis.

Several limitations of this study should be noted. First, the retrospective design limited the analysis to specimens with standard pathological examinations for MVI, while those with unclear MVI reports were excluded. Second, many patients clinically classified as “high-aggressive” did not receive surgical management, potentially introducing selection bias that could affect the comparability of the results. Consequently, the clinical utility of this diagnostic approach requires independent validation in future studies.

In summary, MVI and/or ES grade III-IV is an unfavorable biological character of HCC. We provided a clinically available LTIS that can be alternatively combined with current staging systems for the risk stratification of HCC patients. Using this presurgical tool, it may serve as a pre-treatment triage test to identify patients with high-aggressive HCCs who are potential candidates for wide resection margin, liver transplantation or clinical trials of adjuvant therapy, thus leading to an important step targeting for personalized prognostication.

Author Contribution

Conception and design: ZYD, ZYL, QJRDevelopment of methodology: LQP, LX, YKL, SSW, XX, QK, XJQ, LCY, YHQ, LYY, ZYL, QJR, ZYDAcquisition of data (acquired and managed patients, provided facilities, etc.): LQP, LX, YKL, SSW, XX, YHQ, ZYL, QJR, ZYDAnalysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): LQP, LX, YKL, SSW, XX, QK, XJQ, LCY, YHQWriting, review, and/or revision of the manuscript: ZYD, ZYL, QJR Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): LQP, LX, YKL, SSW, XX, YHQ, YYL, ZYL, QJR, ZYD Study supervision and guarantors: ZYD, ZYL, QJR

Acknowledgement

We thank the First Affiliated Hospital of Nanjing Medical University, the First Affiliated Hospital of Xi’an Jiaotong University and Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital for assistance with histopathology.

Villanueva A. Hepatocellular Carcinoma. N Engl J Med.2019;380(15):1450–62. Epub 2019/04/11. doi: 10.1056/NEJMra1713263.
Yang JD, Heimbach JK. New advances in the diagnosis and management of hepatocellular carcinoma. BMJ.2020;371:m3544. Epub 2020/10/28. doi: 10.1136/bmj.m3544.
European Association for the Study of the Liver. Electronic address eee, European Association for the Study of the L. EASL Clinical Practice Guidelines: Management of hepatocellular carcinoma. J Hepatol.2018;69(1):182–236. Epub 2018/04/10. doi: 10.1016/j.jhep.2018.03.019.
Chan AWH, Zhong J, Berhane S, et al. Development of pre and post-operative models to predict early recurrence of hepatocellular carcinoma after surgical resection. J Hepatol.2018;69(6):1284–93. Epub 2018/09/22. doi:10.1016/j.jhep.2018.08.027.
Banerjee S, Wang DS, Kim HJ, et al. A computed tomography radiogenomic biomarker predicts microvascular invasion and clinical outcomes in hepatocellular carcinoma. Hepatology.2015;62(3):792–800. doi: 10.1002/hep.27877.
Renzulli M, Brocchi S, Cucchetti A, et al. Can Current Preoperative Imaging Be Used to Detect Microvascular Invasion of Hepatocellular Carcinoma? Radiology.2016;279(2):432–42. doi: 10.1148/radiol.2015150998.
Lee S, Kim SH, Lee JE, Sinn DH, Park CK. Preoperative gadoxetic acid-enhanced MRI for predicting microvascular invasion in patients with single hepatocellular carcinoma. J Hepatol.2017;67(3):526–34. Epub 2017/05/10. doi: 10.1016/j.jhep.2017.04.024.
Choi JY, Lee JM, Sirlin CB. CT and MR imaging diagnosis and staging of hepatocellular carcinoma: part II. Extracellular agents, hepatobiliary agents, and ancillary imaging features. Radiology.2014;273(1):30–50. Epub 2014/09/24. doi: 10.1148/radiol.14132362.
Lei Z, Li J, Wu D, et al. Nomogram for Preoperative Estimation of Microvascular Invasion Risk in Hepatitis B Virus-Related Hepatocellular Carcinoma Within the Milan Criteria. JAMA surgery.2016;151(4):356–63. doi: 10.1001/jamasurg.2015.4257.
Xu X, Zhang HL, Liu QP, et al. Radiomic analysis of contrast-enhanced CT predicts microvascular invasion and outcome in hepatocellular carcinoma. J Hepatol.2019;70(6):1133–44. Epub 2019/03/17. doi: 10.1016/j.jhep.2019.02.023.
Johnson PJ, Berhane S, Kagebayashi C, et al. Assessment of liver function in patients with hepatocellular carcinoma: a new evidence-based approach-the ALBI grade. Journal of clinical oncology: official journal of the American Society of Clinical Oncology.2015;33(6):550–8. doi: 10.1200/JCO.2014.57.9151.
Mitchell DG, Bruix J, Sherman M, Sirlin CB. LI-RADS (Liver Imaging Reporting and Data System): summary, discussion, and consensus of the LI-RADS Management Working Group and future directions. Hepatology.2015;61(3):1056–65. doi: 10.1002/hep.27304.
Roayaie S, Blume IN, Thung SN, et al. A system of classifying microvascular invasion to predict outcome after resection in patients with hepatocellular carcinoma. Gastroenterology.2009;137(3):850–5. Epub 2009/06/16. doi: 10.1053/j.gastro.2009.06.003.
Vickers AJ, Van Calster B, Steyerberg EW. Net benefit approaches to the evaluation of prediction models, molecular markers, and diagnostic tests. Bmj.2016;352:i6. doi: 10.1136/bmj.i6.
Chandarana H, Robinson E, Hajdu CH, et al. Microvascular invasion in hepatocellular carcinoma: is it predictable with pretransplant MRI? AJR American journal of roentgenology.2011;196(5):1083–9. doi: 10.2214/AJR.10.4720.
Zhang XP, Chai ZT, Gao YZ, et al. Postoperative adjuvant sorafenib improves survival outcomes in hepatocellular carcinoma patients with microvascular invasion after R0 liver resection: a propensity score matching analysis. HPB : the official journal of the International Hepato Pancreato Biliary Association.2019. Epub 2019/06/04. doi: 10.1016/j.hpb.2019.04.014.
Terrault NA, Lok ASF, McMahon BJ, et al. Update on prevention, diagnosis, and treatment of chronic hepatitis B: AASLD 2018 hepatitis B guidance. Hepatology.2018;67(4):1560–99. doi: 10.1002/hep.29800.
Kudo M, Matsui O, Izumi N, et al. JSH Consensus-Based Clinical Practice Guidelines for the Management of Hepatocellular Carcinoma: 2014 Update by the Liver Cancer Study Group of Japan. Liver cancer.2014;3(3–4):458–68. doi: 10.1159/000343875.
Kamarajah SK, Frankel TL, Sonnenday C, Cho CS, Nathan H. Critical evaluation of the American Joint Commission on Cancer (AJCC) 8th edition staging system for patients with Hepatocellular Carcinoma (HCC): A Surveillance, Epidemiology, End Results (SEER) analysis. Journal of surgical oncology.2018;117(4):644–50. doi: 10.1002/jso.24908.

No competing interests reported.

Download PDF

Editorial decision: Revision requested
04 Oct, 2024
Reviews received at journal
04 Oct, 2024
Reviewers agreed at journal
13 Sep, 2024
Reviews received at journal
11 Sep, 2024
Reviewers agreed at journal
11 Sep, 2024
Reviewers invited by journal
10 Sep, 2024
Editor assigned by journal
02 Sep, 2024
Submission checks completed at journal
02 Sep, 2024
First submitted to journal
01 Sep, 2024

You are reading this latest preprint version

Liver Tumor Imaging Staging: A Multi-Institutional Study of a Preoperative Staging Tool for Hepatocellular Carcinoma

Status:

Version 1

Abstract

Background & Aims:

Methods

Results

Conclusion

Figures

Introduction

Materials and methods

Study Population

Image Analysis

Model Development and Validation

Net Clinical Benefit of LTIS and Postsurgical Prognosis

Statistical Analysis

Results

Baseline characteristics

Radiomic Feature Selection

Prediction Model Development and Validation

Net Clinical Benefit of LTIS Score on Postsurgical Prognosis

Discussion

Declarations

Author Contribution

Acknowledgement

References

Additional Declarations

Supplementary Files

Status:

Version 1