Establishment and Validation of Diagnostic Model of Microvascular Invasion in Hepatocellular Carcinoma

doi:10.21203/rs.3.rs-2618122/v1

Download PDF

Research Article

Establishment and Validation of Diagnostic Model of Microvascular Invasion in Hepatocellular Carcinoma

https://doi.org/10.21203/rs.3.rs-2618122/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background: Microvascular invasion (MVI) is closely related to high recurrence and poor prognosis in patients with hepatocellular carcinoma (HCC). We aimed to establish a preoperative diagnostic model of MVI for HCC based on the clinical characteristics and serum detectors of HCC patients.

Methods: In total, 1027 hepatocellular carcinoma patients hospitalized at Shandong Provincial Hospital from January 2016 to August 2021 were included and randomly divided into the development group and validation group at a ratio of 3:1. Independent sample t test, Mann-Whitney U test, Chi-square test and Fisher exact test were applied to assess the difference of clinical characteristics and serum index results of the two cohorts. Univariate and multivariate logistic regression analyses were carried out to screen the independent risk factors of HCC patients with microvascular invasion. By using these independent risk factors, a preoperative diagnostic nomogram of HCC for MVI was established and verified. The receiver operating characteristic (ROC) curves, calibration curves and decision curve analyses (DCA) were used to estimate the discrimination and clinical application of the nomogram. In addition, the value of this diagnosticmodel in diagnosing microvascular invasion in different stages of hepatocellular carcinoma was further discussed.

Results: Through univariate and multivariate analyses, independent riskfactors for MVI of HCCinvolved Hepatitis B virus infection (HBV), large tumor diameter, higher logarithm of Alpha-fetoprotein (Log AFP), higher logarithm of AFP-L3% (Log AFP-L3%), higher logarithm of protein induced by vitamin K absence or antagonist-II (Log PIVKA-II) and higher logarithm of Carbohydrate antigen 125 (Log CA125). The nomogramincorporating these six independent risk factors was finally established. The areas under the ROC curve (AUC) assessed by the nomogram for MVI of HCC in development cohort and validation cohort were 0.806 (95% CI, 0.773～0.839) and 0.818 (95% CI, 0.763～0.874) respectively. The calibration curve revealed that the judged results for MVI of hepatocellular carcinoma using our established diagnostic model were highly consistent with the postoperative pathological results. The decision curve analysis (DCA) showed promising clinical application of the diagnostic nomogram. Moreover, we also found that the diagnostic model had better application value in hepatocellular carcinoma with higher malignancy.

Conclusion: An effective preoperative diagnostic model for MVI of HCC based on readily available tumor markers and clinical characters has been established, which can be significant and easily implemented for MVI diagnosis.

Hepatocellular carcinoma

Diagnostic model

Microvascular invasion

Primary liver cancer is the sixth most common cancer and the third leading cause of cancer-related death in the whole world in 2020 [1]. Among them, HCC is the most important type of liver cancer (comprising 75%-85%). Postoperative recurrence is a significant risk factor affecting the survival time of HCC patients. Previous studies confirmed that microvascular invasion (MVI) was one of the most important risk factors for early recurrence of HCC [2]. The presence of MVI is a histopathologic feature that indicates aggressive behavior of the HCC, which is closely related to worse prognosis after treatment [3–5]. However, the diagnosis of MVI mainly depends on postoperative pathological examination at present, which is lack of guiding significance for preoperative evaluation and treatment selection of HCC patients. Accurate preoperative diagnosis of the occurrence of MVI for HCC patients can help clinicians select appropriate surgical methods for patients based on the evaluation results of patients. For HCC patients with MVI risk, a wider range of resection should be preferred when removing liver tumors [6]. Therefore, it is of great clinical significance to research the noninvasive diagnostic indicators of microvascular invasion before surgery to guide tumor diagnosis and treatment.

AFP is a glycoprotein, and its elevation is usually related to HCC. According to the binding ability of AFP to lectin lens culinaris agglutin (LCA), AFP can be divided into three different subtypes. Among them, AFP - L3 is the main type of AFP in serum of HCC patients. PIVKA-II, also known as des-γ-carboxyprothrombin (DCP), has been regarded as an ideal biomarker for the diagnosis and evaluation of HCC recently. CA125 is a mucin glycoprotein produced by mucin 16 (MUC16) gene [7]. Many scholars have made great efforts on preoperative evaluation of MVI in recent years. It was reported that PIVKA-II and AFP were highly expressed in the serum of HCC patients with MVI [8]. Pote et al. found that serum PIVKA-II > 90 mAU/ml or AFP > 100 IU were independent risk factors for MVI [9, 10]. Research had found that high preoperative serum CA125 level of HCC patients was significantly associated with microvascular invasion, elevated serum CA125 often predicted poor prognosis of patients with HCC [11, 12]. Large tumor diameter was also significantly associated with the presence of MVI. Some studies showed that it had the higher risk of MVI when the diameter of liver cancer was greater than 3cm [9]. Although previous studies have confirmed that serum tumor markers and tumor characteristics are related to the occurrence of MVI, their clinical value in the preoperative risk assessment of MVI is still unsatisfactory. It is insufficient to rely on the single indicator to predict MVI, and the diagnosis of MVI needs to be combined with multiple key factors.

Given the lack of highly specific and sensitive biomarker to diagnose MVI, the establishment of diagnostic model incorporating several independent risk factors may be a viable way to solve the problem. Nomogram is a simple graphic system recently discovered and widely used, which may be more valuable than traditional methods for earlier diagnosis and prognosis of a wide variety of malignant tumors [13, 14]. The nomogram is a diagnostic or predictive tool based on biologic and clinical variables to graphically depict a statistical model that generates a probability of a clinical event for a given individual, such as cancer occurrence or death. The graphical presentation may be a number of rulers where the key variables are listed separately, with a number of points assigned to the variable. Then, the score obtained by the sum of all the variables is matched to the probability of outcome. With increased accuracy, the nomogram is a commonly used tool to estimate occurrence or prognosis in oncology and medicine [15–17]. A number of diagnostic models for MVI have been developed over recent years. However, these models often include too many variables that are not easy to obtain, which necessarily limit their clinical application. In order to effectively diagnose microvascular invasion before surgery, our study establishes and verifies the nomogram model for the diagnosis of MVI for HCC patients based on readily available clinical characteristics and the results of routine laboratory tests.

Patients

All together 1027 patients with HCC who underwent surgery in Shandong Provincial Hospital from January 2016 to August 2021 were included in this retrospective study. The patients were selected according to the inclusion and exclusion criteria just as shown in Fig. 1. The criteria for inclusion was patient with HCC confirmed by pathological examination over 18 years old. The exclusion reasons were as follows: (1) combined with other tumor; (2) preoperative treatment was administered; (3) puncture biopsy without operation; (4) incomplete information; (5) multiple hepatocellular carcinomas (two or more tumors). Patients who met the criteria were divided into the development group and validation group randomly at a ratio of 3:1. Ethical approval involving human participants was obtained from the Shandong Provincial Hospital Research Ethics Committee. Because the relevant data were anonymous, the requirement for informed consent was waived [18].

Clinicopathologic Variables

The demographic information and clinicopathological variables of participants included age, gender, hepatitis B history, smoking history, history of alcohol intake, past history of hypertension and diabetes, clinicopathological staging of HCC, tumor diameter, etc. Among them, tumor diameter was determined by preoperative magnetic resonance imaging. Routine blood tests included AFP, PIVKA-II, AFP-L3%, CA125, carcinoembryonic antigen (CEA), Carbohydrate antigen 199 (CA199), aspartate transaminase (AST), alanine transaminase (ALT), total bilirubin (TBIL), direct bilirubin (DBIL), γ-glutamyl transpeptidase (GGT), total protein (TP), albumin (ALB), platelet (PLT), white blood cell (WBC) and red blood cell (RBC). These tests we mentioned above were detected within three days before scheduled surgery. AFP and AFP-L3 were measured using immunofluorescence on the automatic electrophoresis fluorescence immunoassay instrument (mTAS Wako i30, Japan). The percentage of AFP-L3 (AFP-L3%) was calculated by AFP-L3 divided by AFP. PIVKA-II was detected by the method of chemiluminescence enzyme immunoassay on the automatic chemiluminescence immunoassay analyzer (LUMIPULSE®G1200, Japan). Serum CA125, CA199 and CEA were detected with automatic electrochemiluminescence analyzer (Roche Diagnostics, Germany). The detections of serum ALT, AST, TBIL, DBIL, GGT, TP and ALB were performed using the automatic biochemical analyzer (AU5831, USA). The blood PLT, RBC and WBC were detected with the fully automated hematology analyzer (Sysmex XN-9000, Japan). All the reagents required above were all original kit, and the tests were carried out in strict accordance with the standard operation procedures [7]. MVI was defined as the invasion of tumor cells in a portal vein, hepatic vein, or a large capsular vessel of the surrounding hepatic tissue. The presence of MVI was the histopathologic feature that could only be diagnosed reliably in resected specimens. All surgical specimens were examined by two experienced pathologists to detect the presence of MVI and Edmondson-Steiner grade.

Statistical Analysis

Numerical variables were expressed as the mean with standard deviation (SD) or the median with interquartile range (IQR), which were compared by student’s t-test or Mann–Whitney test respectively. Categorical variables were expressed as frequencies and Pearson’s χ2 test or Fisher exact test was used to compare their difference. Logarithmic transformation was applied to these variables with skewed distributions such as AFP, PIVKA-II, AFP-L3%, CEA, CA125, CA199, ALT, AST, GGT, TBIL and DBIL [19]. Univariate and multivariate logistic regression analyses were carried out to screen independent risk factors for MVI of HCC patients. Subsequently, the nomogram was drawn by using these independent risk factors. Each risk factor was set the corresponding score according to its importance, and the total score of every HCC patient was calculated using this nomogram, by which we obtained the risk probability of MVI presence finally. The diagnostic value of the nomogram was evaluated by the ROC curve and calibration curve. Besides that, the decision curve analysis (DCA) was used to estimate the clinical application of the nomogram by quantifying net benefits against a range of threshold probabilities. We made further evaluation for the value of this diagnostic model in diagnosing microvascular invasion of different stages of hepatocellular carcinoma. The nomogram was established and validated with R (http://www.R-project.org) and EmpowerStats software (www.empowerstats.com, Boston MA, USA) [20]. IBM SPSS software (version 25.0, USA) and MedCalc software (version 20.0.8, Belgium) were also used to perform the relevant analyses. The P values less than 0.05 were considered to indicate statistical significance.

Demographics and Clinicopathologic Variables of Patient

During research, a total of 1027 consecutive HCC patients who received resection of hepatocellular carcinoma and met the selection criteria were enrolled and randomly divided into the development group and validation group at the ratio of 3:1. Among them, 770 HCC patients formed the training cohort, which included 228 patients with microvascular invasion, and the proportion of MVI was 29.61%. The validation cohort consisted of 257 HCC patients, 73 of whom had microvascular infiltration, accounting for 28.4%. The demographics and clinicopathologic variables of patients in the training and validation cohort were listed in Table 1. A comparison of baseline data showed that most of the variables of patients in the two cohorts had no significant difference. Meanwhile, the baseline clinicopathologic data were compared for HCC patients with or without MVI of training group (Table 2).

Table 1

Baseline characteristics of the development group and validation group
Variables	Training (n = 770)	Validation (n = 257)	P value
Age (years)	57.14 (9.94)	57.33 (9.98)	0.795
Sex			0.186
Male	625 (81.17%)	218 (84.82%)
Female	145 (18.83%)	39 (15.18%)
MVI			0.713
No	542 (70.39%)	184 (71.60%)
Yes	228 (29.61%)	73 (28.40%)
AFP (ng/mL)	39.45 (4.7-536.1)	39.43 (4.20-785.8)	0.946
AFP-L3%	8.6 (0.5–36.9)	9.2 (0.5–37.1)	0.982
CEA (ng/mL)	2.50 (1.69–3.74)	2.80 (1.81–3.65)	0.081
CA125 (U/mL)	12.59 (8.84–21.84)	11.75 (8.97–21.50)	0.446
CA199 (IU/mL)	17.12 (10.79-28)	17.02 (10.90-29.48)	0.592
PIVKA-Ⅱ (mAU/mL)	381.00 (53.76-2943.34)	363.00 (51.82–2925)	0.792
AST (U/L)	36.00 (25.00–54.00)	35.00 (25.00–53.00)	0.676
ALT (U/L)	31.00 (21.00–48.00)	32.00 (21.00–49.00)	0.587
TP (g/L)	70.06 (7.12)	69.54 (7.37)	0.318
ALB (g/L)	39.93 (5.38)	40.13 (5.06)	0.6
GGT (U/L)	51.50 (29.00-102.75)	50.00 (27.00–89.00)	0.166
TBIL (µmol/L)	16.38 (12.30-22.02)	16.30 (12.70-21.63)	0.88
DBIL (µmol/L)	3.60 (2.60–5.29)	3.70 (2.80–5.20)	0.477
PLT (10⁹/L)	173.95 (81.80)	158.91 (80.24)	0.011
RBC (10¹²/L)	4.47 (0.62)	4.39 (0.63)	0.076
WBC (10⁹/L)	5.54 (2.33)	5.27 (2.37)	0.103
Tumor diameter (cm)	5.29 (3.73)	4.92 (3.63)	0.166
HBV			0.061
No	91 (11.82%)	42 (16.34%)
Yes	679 (88.18%)	215 (83.66%)
HBP			0.698
No	581 (75.45%)	197 (76.65%)
Yes	189 (24.55%)	60 (23.35%)
DM			0.412
No	680 (88.31%)	222 (86.38%)
Yes	90 (11.69%)	35 (13.62%)
Smoke			0.649
No	455 (59.09%)	156 (60.70%)
Yes	315 (40.91%)	101 (39.30%)
Drink			0.804
No	489 (63.51%)	161 (62.65%)
Yes	281 (36.49%)	96 (37.35%)
Edmondson-Steiner			1.000
Ⅰ-Ⅱ	561 (73.33%)	187 (73.33%)
Ⅲ-Ⅳ	204 (28.67%)	68 (26.67%)
Note: Categorical variables were expressed as frequency. Continuous variables were expressed as mean (SD) or median with interquartile range (IQR).
Abbreviations: MVI, microvascular invasion; PLT, platelet; RBC, red blood cell; WBC, white blood cell; AFP, α-fetoprotein level; AFP-L3, an isoform of AFP characterized by the presence of an a 1–6– linked residue on the AFP carbohydrate side chain, CA199, Carbohydrate antigen 199; CEA, carcinoembryonic antigen; PIVKA-II, protein induced by vitamin K absence or antagonist-II; CA125, Carbohydrate antigen 125; ALT, alanine transaminase; AST, aspartate transaminase; GGT, gamma glutamyl transpeptidase; TP, total protein; ALB, albumin; TBIL, total bilirubin; DBIL, direct bilirubin; HBV, hepatitis B virus; HBP, high blood pressure; DM, diabetes mellitus.

Table 2

Demographic information and clinicopathological characteristics of training cohort
Variables	MVI - (n = 542)	MVI + (n = 228)	P value
Age (years)	57.60 (9.91)	56.05 (9.94)	0.048
Sex			0.427
Male	436 (80.44%)	189 (82.89%)
Female	106 (19.56%)	39 (17.11%)
AFP (ng/mL)	17.13 (3.70-197.07)	270.30 (19.98–2229.00)	< 0.001
AFP-L3%	5.25 (0.50-23.38)	26.45 (5.77–60.62)	< 0.001
CEA (ng/mL)	2.52 (1.72–3.74)	2.44 (1.60–3.75)	0.693
CA125 (U/mL)	11.61 (8.39–18.53)	17.59 (10.95–29.74)	< 0.001
CA199 (IU/mL)	17.80 (11.00-28.82)	15.37 (10.08–27.77)	0.456
PIVKA-Ⅱ (mAU/mL)	180.77 (38.70-1252.44)	1770.0 (249.86-14538.56)	< 0.001
AST (U/L)	33.00 (24.00–50.00)	41.00 (29.00-61.25)	< 0.001
ALT (U/L)	29.00 (20.25-47.00)	33.00 (21.00–50.00)	0.102
TP (g/L)	70.00 (7.31)	70.20 (6.64)	0.733
ALB (g/L)	40.07 (5.53)	39.58 (4.98)	0.247
GGT (U/L)	45.50 (26.25–96.75)	72.00 (41.00-141.50)	< 0.001
TBIL (µmol/L)	16.17 (12.15–22.16)	16.70 (12.80-21.66)	0.365
DBIL (µmol/L)	3.44 (2.60–5.29)	3.98 (2.86–5.23)	0.013
PLT (10⁹/L)	168.24 (72.46)	187.53 (99.47)	0.003
RBC (10¹²/L)	4.48 (0.62)	4.47 (0.62)	0.912
WBC (10⁹/L)	5.44 (2.21)	5.79 (2.58)	0.057
Tumor diameter (cm)	4.31 (3.09)	7.61 (4.10)	< 0.001
HBV			0.015
No	74 (13.65%)	17 (7.46%)
Yes	468 (86.35%)	211 (92.54%)
HBP			0.144
No	401 (73.99%)	180 (78.95%)
Yes	141 (26.01%)	48 (21.05%)
DM			0.74
No	480 (88.56%)	200 (87.72%)
Yes	62 (11.44%)	28 (12.28%)
Smoke			0.448
No	325 (59.96%)	130 (57.02%)
Yes	217 (40.04%)	98 (42.98%)
Drink			0.066
No	333 (61.44%)	156 (68.42%)
Yes	209 (38.56%)	72 (31.58%)
Edmondson-Steiner			< 0.001
Ⅰ-Ⅱ	463 (86.06%)	98 (43.17%)
Ⅲ-Ⅳ	75 (13.94%)	129 (56.83%)
Note: Categorical variables were expressed as frequency. Continuous variables were expressed as mean (SD) or median with interquartile range (IQR).
Abbreviations: MVI, microvascular invasion; PLT, platelet; RBC, red blood cell; WBC, white blood cell; AFP, α-fetoprotein level; AFP-L3, an isoform of AFP characterized by the presence of an a 1–6– linked residue on the AFP carbohydrate side chain, CA199, Carbohydrate antigen 199; CEA, carcinoembryonic antigen; PIVKA-II, protein induced by vitamin K absence or antagonist-II; CA125, Carbohydrate antigen 125; ALT, alanine transaminase; AST, aspartate transaminase; GGT, gamma glutamyl transpeptidase; TP, total protein; ALB, albumin; TBIL, total bilirubin; DBIL, direct bilirubin; HBV, hepatitis B virus; HBP, high blood pressure; DM, diabetes mellitus.

Univariate And Multivariate Analysis Of Independent Risk Factors For Mvi Of Hcc Patients

The risk factors that might affect the occurrence of MVI in HCC patients were screened through the univariate analysis of training cohort, and then multivariate logistic regression was conducted on these potential risk factors. As shown in Table 3, the independent risk factors for MVI of HCC involved HBV infection (OR = 2.24, 95% CI: 1.16 ~ 4.33, P = 0.0163), large tumor diameter (OR = 1.19, 95% CI: 1.12 ~ 1.26, P < 0.0001) and the increase of the logarithm of four tumor markers: Log AFP (OR = 1.32, 95% CI: 1.12 ~ 1.56, P = 0.0009), Log AFP-L3% (OR = 1.51, 95% CI: 1.16 ~ 1.96, P = 0.002), Log PIVKA-II (OR = 1.23, 95% CI: 1.01 ~ 1.49, P = 0.0351) and Log CA125 (OR = 1.64, 95% CI: 1.02 ~ 2.65, P = 0.0403).

Table 3

Univariate and multivariate logistic regression analysis of MVI presence based on preoperative data in training cohort
Variables	Univariable		Multivariable
Variables	OR (95%CI)	P value	OR (95%CI)	P value
Age (years)	0.98 (0.97, 1.00)	0.0484	N/A
Sex		0.4272
Male	1		N/A
Female	0.85 (0.57, 1.27)		N/A
Log AFP (ng/mL)	1.80 (1.58, 2.05)	< 0.0001	1.32 (1.12, 1.56)	0.0009
Log AFP-L3%	2.28 (1.85, 2.81)	< 0.0001	1.51 (1.16, 1.96)	0.002
Log CEA (ng/mL)	1.03 (0.62, 1.73)	0.9021	N/A
Log CA125 (U/mL)	2.89 (1.91, 4.39)	< 0.0001	1.64 (1.02, 2.65)	0.0403
Log CA199 (IU/mL)	0.96 (0.70, 1.33)	0.8224	N/A
Log PIVKA-Ⅱ (mAU/mL)	2.01 (1.72, 2.35)	< 0.0001	1.23 (1.01, 1.49)	0.0351
Log AST (U/L)	2.22 (1.31, 3.77)	0.0031	N/A
Log ALT (U/L)	1.33 (0.83, 2.13)	0.2378	N/A
TP (g/L)	1.00 (0.98, 1.03)	0.7323	N/A
ALB (g/L)	0.98 (0.96, 1.01)	0.2465	N/A
Log GGT (U/L)	2.94 (1.99, 4.35)	< 0.0001	N/A
Log TBIL (µmol/L)	1.30 (0.72, 2.33)	0.3828	N/A
Log DBIL (µmol/L)	1.52 (0.98, 2.37)	0.0645	N/A
PLT (10⁹/L)	1.00 (1.00, 1.00)	0.0033	N/A
RBC (10¹²/L)	0.99 (0.77, 1.27)	0.9121	N/A
WBC (10⁹/L)	1.06 (1.00, 1.13)	0.0589	N/A
Tumor diameter (cm)	1.28 (1.22, 1.35)	< 0.0001	1.19 (1.12, 1.26)	<0.0001
HBV		0.0166		0.0163
No	1		1
Yes	1.96 (1.13, 3.41)		2.24 (1.16, 4.33)
HBP		0.1448
No	1		N/A
Yes	0.76 (0.52, 1.10)		N/A
DM		0.7401
No	1		N/A
Yes	1.08 (0.67, 1.74)		N/A
Smoke		0.448
No	1		N/A
Yes	1.13 (0.83, 1.54)		N/A
Drink		0.0666
No	1		N/A
Yes	0.74 (0.53, 1.02)		N/A
Abbreviations: PLT, platelet; RBC, red blood cell; WBC, white blood cell; AFP, α-fetoprotein level; AFP-L3, an isoform of AFP characterized by the presence of an a 1–6– linked residue on the AFP carbohydrate side chain, CA199, Carbohydrate antigen 199; CEA, carcinoembryonic antigen; PIVKA-II, protein induced by vitamin K absence or antagonist-II; CA125, Carbohydrate antigen 125; ALT, alanine transaminase; AST, aspartate transaminase; GGT, gamma glutamyl transpeptidase; TP, total protein; ALB, albumin; TBIL, total bilirubin; DBIL, direct bilirubin; HBV, hepatitis B virus; HBP, high blood pressure; DM, diabetes mellitus.

Development And Validation Of The Diagnostic Nomogram For Mvi Of Hcc Patients

The independent risk factors of MVI were further incorporated to establish the nomogram for predicting MVI occurrence (Fig. 2). The nomogram can be better to predict the occurrence of MVI with the AUC of 0.806 (95% CI: 0.773ཞ0.839) and 0.818 (95% CI: 0.763ཞ0.874) in the development group and validation group respectively (Fig. 3). The sensitivities, specificities, positive and negative predictive values of the nomogram were also calculated and listed Table 4. The calibration curves showed that the MVI of HCC diagnosed by the nomogram were highly consistent with the actual results (Fig. 4). Considering the clinical benefits and patient preferences, DCA curve displayed that the nomogram would offer net benefits over the “treat-all” or “treat-none” groups. After verification, using this nomograph to predict the risk of MVI in HCC showed greater benefit (Fig. 5).

Table 4

The value of the diagnostic model established in this study in predicting the occurrence of MVI
Variables	Training cohort	Validation cohort
AUC (95% CI)	0.806(0.773–0.839)	0.818(0.763–0.874)
Specificity	81.00%	73.91%
Sensitivity	67.98%	79.45%
PPV	60.08%	54.72%
NPV	85.74%	90.07%
PLR	3.58	3.05
NLR	0.40	0.28
Abbreviations: MVI, microvascular invasion; AUC, area under the receiver operating characteristic curve; NPV, negative predictive value; PPV, positive predictive value; PLR, positive likelihood ratio; NLR, negative likelihood ratio.

Influence Of Differentiation Degree Of Hepatocellular Carcinoma On The Diagnostic Model

According to Edmondson-Steiner classification of hepatocellular carcinoma, the participants were divided into well differentiated group (including I and II) and poor differentiated group (including III and IV). In order to evaluate the influence of differentiation degree of hepatocellular carcinoma on the diagnostic model for MVI, we compared the difference of the AUC, sensitivity and specificity of the model to diagnose MVI for HCC with well and poor differentiated group. As shown in Table 5, The AUCs (95% CI) of the diagnostic model of MVI in well differentiated and poor differentiated HCC were 0.729 (0.683 ~ 0.774) and 0.846 (0.796 ~ 0.895) respectively. Then the difference of AUC between the two group was compared using the Z-test. It revealed that the difference of AUC was statistically significant, indicating that the differentiation of HCC had an impact on the value of diagnostic model for MVI and the value of diagnostic model in poorly differentiated HCC was significantly higher than that in well differentiated HCC.

Table 5

The value of the diagnostic model established in this study in predicting the occurrence of MVI for different stages of liver cancer
Variables	Well differentiated	Poor differentiated	Z value	P value
AUC (95% CI)	0.729(0.683–0.774)	0.846(0.796–0.895)	3.444	0.0006
Specificity	51.93%	71.43%
Sensitivity	81.75%	86.21%
PPV	25.62%	84.27%
NPV	93.35%	74.47%
PLR	1.70	3.02
NLR	0.35	0.19
Abbreviations: MVI, microvascular invasion; AUC, area under the receiver operating characteristic curve; NPV, negative predictive value; PPV, positive predictive value; PLR, positive likelihood ratio; NLR, negative likelihood ratio.

The microvascular invasion of HCC was a typical manifestation of its invasiveness and main factor affecting the prognosis of HCC patients, which also influenced the choices of treatment strategy for HCC [21]. However, no specific diagnostic indicators for preoperative diagnosis of MVI have been found.

This study retrospectively analyzed the clinical data and preoperative blood test results of 1027 patients with hepatocellular carcinoma who underwent hepatectomy in our hospital from January 2016 to August 2021. After univariate and multiple regression analysis, we found that HBV infection, large tumor diameter, elevated Log AFP, Log AFP L3%, Log CA125 and Log PIVKA-II were independent risk factors for MVI. Based on these factors, a preoperative diagnostic model of MVI in HCC was established and validated. The AUC (95% CI) of the diagnostic model for predicting MVI in HCC in the development group and validation group was 0.806 (95% CI, 0.773ཞ0.839) and 0.818 (95% CI, 0.763ཞ0.874), respectively. It was indicated that the diagnostic model had a high predictive value for the occurrence of MVI in HCC. Moreover, the calibration curve also showed that using the nomogram to diagnose MVI of HCC had a high consistency with the actual pathological results. The previous diagnosis models lacked the evaluation of their clinical application value. DCA could remedy this defect, which was an effective method to evaluate the clinical application prospect [22–24]. The results of DCA curve showed that using our nomogram to predict the occurrence of MVI had high clinical value. In this study, we also evaluated the value difference of the diagnostic model in diagnosing MVI for HCC with different malignancies. It was found that the higher the malignancy of the tumor, the greater the value of using our diagnostic model to predict the occurrence of MVI.

Hepatitis B virus infection was the main risk factor of hepatocellular carcinoma. Previous studies showed that high serum deoxyribonucleic acid (DNA) of HBV before operation was an independent risk factor for MVI in HCC patients [25, 26]. In addition, the enhanced level of AFP and PIVKA-II were also independent dangerous factors for MVI of HCC [10], which was basically consistent with the relevant results of our research. However, it was difficult to achieve accurate prediction of MVI just using single indicator, which was necessary to combine multiple factors to provide a more clinically valuable way. Shirabe et al proposed a scoring system including tumor size, serum concentration of PIVKA-II and the degree of tumor differentiation to predict MVI of HCC [27, 28]. However, this evaluation system required tumor biopsy, which had side effects on patients. Preoperative imaging examination could obtain data such as tumor diameter, tumor number, tumor shape and so on, which might be the factors related to MVI [29, 30]. If the blood test results could be combined with imaging examination to predict MVI, the effect may be better. A study based on the clinical data of 1004 cases of hepatitis B virus -related HCC found that the preoperative risk factors of MVI included large tumor diameter, multiple tumor nodules, incomplete tumor envelope, AFP > 20 ng/ml, platelet < 100 × 10³/µl, hepatitis B virus DNA > 10⁴ IU/ml in peripheral blood and a typical dynamic pattern of tumors on enhanced magnetic resonance imaging (MRI). Incorporating these seven factors, the diagnostic model was established and achieved good concordance in predicting MVI [13]. The screened risk factors of the above model were mainly from the results of preoperative blood tests and imaging examinations, which was similar to our diagnostic model. But the risk factors we screened were all from preoperative routine examinations, which were easier to obtain. In addition, we can judge the tumor diameter only with ordinary MRI without enhanced MRI, which laid a foundation for the extensive clinical application of this diagnostic model. We further carried out logarithmic conversion of blood test results among risk factors, which could achieve quantitative statistics. The AUC of our diagnostic model in predicting MVI were all greater than 0.8 in the training and validation cohorts. Therefore, the diagnostic model had the advantages of simplicity and efficiency.

Of course, our research also had some limitations: since the data of this research were obtained from one single institution, which needed further verification by external and large samples.

In conclusion, we demonstrated that HBV infection, large tumor diameter, higher Log AFP, Log PIVKA-II, Log AFP-L3% and Log CA125 were the independent risk factors for MVI of HCC. The nomogram incorporating these six factors achieved an optimal preoperative prediction of MVI in HCC, which can be helpful for the assessment of HCC patient and the choice of rational therapeutic strategy.

Ethics statement

This study was conducted in accordance with the ethical standards of the Declaration of Helsinki and approved by Ethics Committee of the Shandong Provincial Hospital Affiliated to Shandong First Medical University (Approval Number: SWXY: NO.2021-461). Because the relevant data were anonymous, the requirement for informed consent was waived by Ethics Committee of the Shandong Provincial Hospital Affiliated to Shandong First Medical University.

Consent to Publication

Not applicable.

Data Availability statement

The datasets used and/or analyzed during this research can be obtained from the corresponding author upon reasonable request.

Conflict of interest

The authors declare that they have no competing interests.

Funding

None.

Acknowledgment

We appreciated the contributions of Department of Hepatobiliary Surgery, Shandong Provincial Hospital Affiliated to Shandong First Medical.

Authors’ contributions

YS conceived the research and designed the experiments. DH, YF, CP, NZ performed the data acquisition and statistical analysis; XW drafted and revised the manuscript. All authors read and approved the submitted version of the manuscript.

Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, Bray F. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209–49.
Torzilli G, Donadon M, Belghiti J, Kokudo N, Takayama T, Ferrero A, Nuzzo G, Vauthey JN, Choti MA, De Santibanes E, et al. Predicting Individual Survival After Hepatectomy for Hepatocellular Carcinoma: a Novel Nomogram from the "HCC East & West Study Group. J Gastrointest surgery: official J Soc Surg Aliment Tract. 2016;20(6):1154–62.
Erstad DJ, Tanabe KK. Prognostic and Therapeutic Implications of Microvascular Invasion in Hepatocellular Carcinoma. Annals of surgical oncology 2019, 26(5):1474–1493.
Rodríguez-Perálvarez M, Luong TV, Andreana L, Meyer T, Dhillon AP, Burroughs AK. A systematic review of microvascular invasion in hepatocellular carcinoma: diagnostic and prognostic variability. Ann Surg Oncol. 2013;20(1):325–39.
Zhang X, Li J, Shen F, Lau WY. Significance of presence of microvascular invasion in specimens obtained after surgical treatment of hepatocellular carcinoma.Journal of gastroenterology and hepatology 2018, 33(2):347–354.
Hirokawa F, Hayashi M, Miyamoto Y, Asakuma M, Shimizu T, Komeda K, Inoue Y, Uchiyama K. Outcomes and predictors of microvascular invasion of solitary hepatocellular carcinoma. Hepatol research: official J Japan Soc Hepatol. 2014;44(8):846–53.
Si YQ, Wang XQ, Pan CC, Wang Y, Lu ZM. An Efficient Nomogram for Discriminating Intrahepatic Cholangiocarcinoma From Hepatocellular Carcinoma: A Retrospective Study.Frontiers in oncology 2022, 12:833999.
Eguchi S, Takatsuki M, Hidaka M, Soyama A, Tomonaga T, Muraoka I, Kanematsu T. Predictor for histological microvascular invasion of hepatocellular carcinoma: a lesson from 229 consecutive cases of curative liver resection. World J Surg. 2010;34(5):1034–8.
McHugh PP, Gilbert J, Vera S, Koch A, Ranjan D, Gedaly R. Alpha-fetoprotein and tumour size are associated with microvascular invasion in explanted livers of patients undergoing transplantation with hepatocellular carcinoma. HPB: the official journal of the International Hepato Pancreato Biliary Association. 2010;12(1):56–61.
Poté N, Cauchy F, Albuquerque M, Voitot H, Belghiti J, Castera L, Puy H, Bedossa P, Paradis V. Performance of PIVKA-II for early hepatocellular carcinoma diagnosis and prediction of microvascular invasion. J Hepatol. 2015;62(4):848–54.
Zhou S, Wang Z, Li M, Wu L. Elevated Preoperative Serum CA125 Predicts Larger Tumor Diameter in Patients with Hepatocellular Carcinoma and Low AFP Levels.BioMed research international 2019, 2019:6959637.
Huang Y, Zeng J, Liu T, Lin X, Guo P, Zeng J, Zhou W, Liu J. Prognostic Significance of Elevated Preoperative Serum CA125 Levels After Curative Hepatectomy for Hepatocellular Carcinoma. OncoTargets and therapy. 2020;13:4559–67.
Lei Z, Li J, Wu D, Xia Y, Wang Q, Si A, Wang K, Wan X, Lau WY, Wu M, et al. Nomogram for Preoperative Estimation of Microvascular Invasion Risk in Hepatitis B Virus-Related Hepatocellular Carcinoma Within the Milan Criteria. JAMA Surg. 2016;151(4):356–63.
Zou Q, Li J, Wu D, Yan Z, Wan X, Wang K, Shi L, Lau WY, Wu M, Shen F. Nomograms for Pre-operative and Post-operative Prediction of Long-Term Survival of Patients Who Underwent Repeat Hepatectomy for Recurrent Hepatocellular Carcinoma. Ann Surg Oncol. 2016;23(8):2618–26.
Balachandran VP, Gonen M, Smith JJ, DeMatteo RP. Nomograms in oncology: more than meets the eye. Lancet Oncol. 2015;16(4):e173–180.
Caulfield S, Menezes G, Marignol L, Poole C. Nomograms are key decision-making tools in prostate cancer radiation therapy. Urol Oncol. 2018;36(6):283–92.
Morlacco A, Modonutti D, Motterle G, Martino F, Dal Moro F, Novara G. Nomograms in Urologic Oncology: Lights and Shadows.Journal of clinical medicine 2021, 10(5).
Filion KB, Azoulay L, Platt RW, Dahl M, Dormuth CR, Clemens KK, Hu N, Paterson JM, Targownik L, Turin TC et al. A Multicenter Observational Study of Incretin-based Drugs and Heart Failure.The New England journal of medicine 2016, 374(12):1145–1154.
Pocock SJ, Ferreira JP, Gregson J, Anker SD, Butler J, Filippatos G, Gollop ND, Iwata T, Brueckmann M, Januzzi JL et al. Novel biomarker-driven prognostic models to predict morbidity and mortality in chronic heart failure: the EMPEROR-Reduced trial.European heart journal 2021, 42(43):4455–4464.
Wang K, Zuo P, Liu Y, Zhang M, Zhao X, Xie S, Zhang H, Chen X, Liu C. Clinical and Laboratory Predictors of In-hospital Mortality in Patients With Coronavirus Disease-2019: A Cohort Study in Wuhan, China.Clinical infectious diseases: an official publication of the Infectious Diseases Society of America 2020, 71(16):2079–2088.
Wang G, Jian W, Cen X, Zhang L, Guo H, Liu Z, Liang C, Zhou W. Prediction of Microvascular Invasion of Hepatocellular Carcinoma Based on Preoperative Diffusion-Weighted MR Using Deep Learning. Academic radiology 2021, 28 Suppl 1:S118-s127.
Sande SZ, Li J, D'Agostino R, Yin Wong T, Cheng CY. Statistical inference for decision curve analysis, with applications to cataract diagnosis. Stat Med. 2020;39(22):2980–3002.
Vickers AJ, van Calster B, Steyerberg EW. A simple, step-by-step guide to interpreting decision curve analysis. Diagn Progn Res. 2019;3:18.
Zhang Z, Rousson V, Lee WC, Ferdynus C, Chen M, Qian X, Guo Y. Decision curve analysis: a technical note. Ann Transl Med. 2018;6(15):308.
Yang F, Ma L, Yang Y, Liu W, Zhao J, Chen X, Wang M, Zhang H, Cheng S, Shen F, et al. Contribution of Hepatitis B Virus Infection to the Aggressiveness of Primary Liver Cancer: A Clinical Epidemiological Study in Eastern China. Front Oncol. 2019;9:370.
Chen C, Chen DP, Gu YY, Hu LH, Wang D, Lin JH, Li ZS, Xu J, Wang G. Vascular invasion in hepatitis B virus-related hepatocellular carcinoma with underlying cirrhosis: possible associations with ascites and hepatitis B viral factors? Tumour biology: the journal of the International Society for Oncodevelopmental Biology and Medicine 2015, 36(8):6255–6263.
Gomaa AI, Khan SA, Leen EL, Waked I, Taylor-Robinson SD. Diagnosis of hepatocellular carcinoma.World journal of gastroenterology 2009, 15(11):1301–1314.
Shirabe K, Itoh S, Yoshizumi T, Soejima Y, Taketomi A, Aishima S, Maehara Y. The predictors of microvascular invasion in candidates for liver transplantation with hepatocellular carcinoma-with special reference to the serum levels of des-gamma-carboxy prothrombin. J Surg Oncol. 2007;95(3):235–40.
Jiang HY, Chen J, Xia CC, Cao LK, Duan T, Song B. Noninvasive imaging of hepatocellular carcinoma: From diagnosis to prognosis.World journal of gastroenterology 2018, 24(22):2348–2362.
Xu X, Zhang HL, Liu QP, Sun SW, Zhang J, Zhu FP, Yang G, Yan X, Zhang YD, Liu XS. Radiomic analysis of contrast-enhanced CT predicts microvascular invasion and outcome in hepatocellular carcinoma. J Hepatol. 2019;70(6):1133–44.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Establishment and Validation of Diagnostic Model of Microvascular Invasion in Hepatocellular Carcinoma

Status:

Version 1

Abstract

Figures

Introduction

Materials And Methods

Patients

Clinicopathologic Variables

Statistical Analysis

Results

Demographics and Clinicopathologic Variables of Patient

Univariate And Multivariate Analysis Of Independent Risk Factors For Mvi Of Hcc Patients

Development And Validation Of The Diagnostic Nomogram For Mvi Of Hcc Patients

Influence Of Differentiation Degree Of Hepatocellular Carcinoma On The Diagnostic Model

Discussion

Conclusions

Declarations

References

Additional Declarations

Status:

Version 1