1. Introduction

Hepatocellular carcinoma (HCC) is one of the most common and aggressive human malignancies in the world and the third leading cause of cancer-related death worldwide [1]. The prognosis of HCC patients remains unsatisfactory due to the extremely high incidence of recurrence after curative resection [2, 3]. Currently, the universal blood biomarker used for the risk assessment and surveillance of HCC is serum α-fetoprotein (AFP), but nearly 30% of HCC patients have a normal serum level of AFP (AFP < 25 ng/mL) [4–7]. Data regarding the characteristics and outcomes of patients with tumors that do not produce AFP are limited. Finding another reliable biomarker and identifying factors that allow better stratification of AFP-negative patients with worse outcomes would be beneficial in clinical practice.

Gamma-glutamyltransferase (GGT) is found predominantly on the surface of secretory epithelial cells [8]. Abnormal GGT expression has been reported to play a role in tumor progression, invasion, and anticancer drug resistance [9, 10] and is associated with worse prognosis in several human tumors, including ovarian cancer [11], cervical cancer [12], and endometrial cancer [13]. Large epidemiological studies revealed that high serum levels of GGT are associated with an increased risk of liver cancer [14, 15]. Elevated GGT levels have been reported to be a predictor of poor prognosis in HCC patients [16–19]. However, further analysis of the utility of preoperative levels of GGT in AFP-negative HCC patients is lacking.

Therefore, the specific aim of our study was to evaluate the prognostic value of preoperative GGT in AFP-negative HCC patients after resection.

2. Materials and Methods

This study protocol was conducted in accordance with the ethical guidelines of the 1975 Declaration of Helsinki, and the procedure was approved by the institutional review board of the Sun Yat-sen University Cancer Center. Informed consent was obtained from each patient included in the study. The study was censored on 31 December 2017.

3. Results

3.1. Patient Characteristics

As the cutoff points for preoperative GGT varied greatly in previous studies, ranging from 17.9 to 165 U/mL and to avoid an arbitrary cutoff point, all patients were divided into the following three subgroups using X-tile v.3.6.1 software (Yale University, New Haven, CT): group 1 was composed of patients with preoperative levels of GGT < 37.2 U/L (n = 211, 31.1%), group 2 was composed of patients with preoperative levels of GGT ≥ 37.2 and <102.8 U/L (n = 320, 47.2%), and group 3 was composed of patients with preoperative levels of GGT ≥ 102.8 U/L (n = 147, 21.7%) (Figures 1(a) and 1(b)). The clinicopathologic characteristics are summarized in Table 1. Elevated preoperative serum GGT levels were associated with unfavorable tumor burden, including larger tumor size, multiple tumor number, and microvascular invasion (MVI) (Table 1). A total of 609 (89.8%) patients were classified as BCLC grade A, while 45 (6.6%) patients were classified as grade B, and 24 (3.6%) patients were classified as grade C. At the time of censoring, 219 patients (32.3%) had died of HCC, including 32 patients (15.2%) in group 1, 86 patients (26.9%) in group 2, and 67 patients (45.6%) in group 3. During the follow-up period, HCC recurrence was identified in 68 patients (32.2%) in group 1, 161 patients (50.3%) in group 2, and 83 patients (56.5%) in group 3.

Table 1. Clinical characteristics of patients.

Variables	GGT < 37.2 U/L  (n = 211)	GGT ≤ 37.2 U/L and GGT < 102.8 U/L (n = 320)	GGT ≥ 102.8 U/L (n = 147)	P value
Age (years)	52.0 ± 12.3	53.4 ± 10.8	50.7 ± 11.2	0.060
Sex				<0.001
Male	181 (85.8)	305 (95.3)	142 (96.6)
Female	30 (14.2)	15 (4.7)	5 (3.4)
Cause of disease				0.702
HBV	179 (84.8)	278 (86.9)	126 (85.7)
HCV	2 (0.9)	7 (2.2)	4 (2.7)
Others	31 (14.7)	40 (12.5)	19 (12.9)
Platelet count (109/L)	173.9 ± 58.5	175.4 ± 68.1	183.1 ± 89.5	0.448
Liver cirrhosis				0.167
Present	112 (53.1)	195 (60.9)	89 (60.5)
Absent	99 (46.9)	125 (39.1)	58 (39.5)
Albumin (g/L)	42.5 ± 3.3	41.9 ± 4.0	41.5 ± 4.2	0.051
Serum total bilirubin (μmol/L)	13.9 ± 5.6	14.2 ± 5.8	14.8 ± 8.5	0.424
ALBI grade
1	183 (86.7)	254 (79.4)	115 (78.2)	0.055
2	28 (13.3)	66 (20.6)	32 (21.8)
Tumor size (cm)	4.3 ± 2.3	5.4 ± 2.7	6.7 ± 3.5	<0.001
Tumor number				0.009
≤1	190 (90.0)	263 (82.2)	116 (78.9)
>1	21 (10.0)	57 (17.8)	31 (21.1)
Tumor extent				0.119
Unilobar	206 (97.6)	303 (94.7)	137 (93.2)
Bilobar	5 (2.4)	17 (5.3)	10 (6.8)
Microvascular invasion				0.001
Absent	189 (89.6)	268 (83.8)	109 (74.1)
Present	22 (10.4)	52 (16.2)	38 (25.9)
Surgical time (min)	165.6 ± 60.6	169.5 ± 52.9	179.5 ± 55.6	0.065
Type of hepatectomy†				0.107
Major	36 (17.1)	61 (19.1)	38 (25.9)
Minor	175 (82.9)	259 (80.9)	109 (74.1)
Surgical margin‡ (cm)				0.062
<1	73 (34.6)	136 (42.5)	68 (46.3)
≥1	138 (65.4)	184 (57.5)	79 (53.7)
Intraoperative blood loss (mL)				<0.001
<400	160 (75.8)	217 (67.8)	71 (48.3)
≥400	51 (24.2)	103 (32.2)	76 (51.7)
Encapsulation				0.414
Complete	104 (49.3)	149 (46.6)	62 (42.2)
Incomplete	107 (50.7)	171 (53.4)	85 (57.8)
Tumor differentiation				0.235
Well	48 (22.7)	57 (17.8)	24 (16.3)
Moderate or poor	163 (77.3)	263 (82.2)	123 (83.7)
BCLC stage				<0.001
A	200 (94.8)	288 (90.0)	121 (82.3)
B	10 (4.7)	27 (8.4)	8 (5.4)
C	1 (0.5)	5 (1.6)	18 (12.3)

• Variables are expressed as no. (%). HBV: hepatitis B virus; HCV: hepatitis C virus; ALBI grade: albumin-bilirubin grade; BCLC: Barcelona Clinic Liver Cancer. ^†Major liver resection: resection with more than two lobes; minor liver resection: resection with no more than two lobes. ^‡Surgical margin: the shortest measured distance from the edge of the tumor to the plane of liver transection.

3.2. Significant Factors Affecting Overall Survival

The 1-, 3-, and 5-year OS rates were 97.2%, 75.4%, and 44.5% in group 1; 92.5%, 69.1%, and 39.7% in group 2; and 89.1%, 55.8%, and 27.2% in group 3, respectively (Figure 1(c)) (all P < 0.001). Univariate analysis of OS is shown in Table 2, and parameters with P < 0.05, including alanine aminotransferase (ALT) level, total bilirubin level (TBIL), albumin level, GGT level, ALBI grade, liver cirrhosis, tumor size, tumor number, and MVI, were included in multivariate analysis using Cox proportional hazards regression analysis. Only the TBIL level, GGT level, ALBI grade, tumor number, and MVI remained significant predictors of OS (Figure 2).

Table 2. Univariate analysis of risk factors for overall survival and disease-free survival.

Variables	OS		DFS
	HR (95% CI)	P	HR (95% CI)	P
Age (y)
≤50	Reference		Reference
>50	1.19 (0.88-1.61)	0.251	1.09 (0.87-1.37)	0.458
Gender
Female	Reference		Reference
Male	1.03 (0.60-1.78)	0.905	0.83 (0.53-1.30)	0.415
Etiology
Other	Reference		Reference
HBV/HCV	1.37 (0.85-2.21)	0.192	1.91 (1.28-2.85)	0.002
PLT (109/L)
>100	Reference		Reference
≤100	1.35 (0.88-2.05)	0.166	1.28 (0.92-1.79)	0.149
ALT (U/L)
≤40	Reference		Reference
>40	1.37 (1.03-1.83)	0.032	1.47 (1.18-1.84)	0.001
TBIL (μmol/L)
≤17.2	Reference		Reference
>17.2	1.39 (1.00-1.90)	0.044	1.19 (0.93-1.53)	0.174
Albumin (g/L)
>35	Reference		Reference
≤35	3.45 (1.70-7.02)	0.001	0.84 (0.31-2.24)	0.720
GGT (U/L)
≤37.2	Reference		Reference
37.2-102.8	1.94 (1.29-2.91)	0.001	1.82 (1.37-2.42)	<0.001
≥102.8	3.92 (2.57-5.99)	<0.001	2.53 (1.84-3.49)	<0.001
ALBI grade
1	Reference		Reference
2	1.92 (1.39-2.67)	<0.001	1.32 (1.00-1.75)	0.048
Liver cirrhosis
No	Reference		Reference
Yes	1.41 (1.04-1.92)	0.027	1.22 (0.97-1.54)	0.086
Tumor size, cm
≤5	Reference		Reference
>5	1.84 (1.37-2.47)	<0.001	1.61 (1.28-2.00)	<0.001
Tumor number
≤1	Reference		Reference
>1	3.20 (2.35-4.36)	<0.001	2.01 (1.59-2.72)	<0.001
MVI
Absent	Reference		Reference
Present	2.61 (1.86-3.67)	<0.001	1.96 (1.48-2.59)	<0.001

• OS: overall survival; DFS: disease-free survival; PLT: platelet; ALT: alanine aminotransferase; TBIL: total bilirubin; GGT: gamma-glutamyltransferase; ALBI grade: albumin-bilirubin grade; MVI: microvascular invasion.

3.3. Significant Factors Affecting Disease-Free Survival

The 1-, 3-, and 5-year DFS rates were 88.2%, 59.2%, and 33.6% in group 1; 74.7%, 49.4%, and 24.7% in group 2; and 66.6%, 34.0%, and 15.6% in group 3, respectively (Figure 1(d)) (all P < 0.001). In univariate analysis of DFS, significant differences were observed in the following variables: HBV/HCV, ALT level, GGT level, ALBI grade, tumor size, tumor number, and MVI (Table 2). The results of the multivariate analysis using the Cox proportional hazards regression analysis are shown in Figure 3. HBV/HCV, GGT level, tumor size, tumor number, and MVI remained significant independent prognostic indicators for DFS.

3.4. Subgroup Analysis according to Milan Criteria and Microvascular Invasion

As an increased preoperative GGT level was associated with unfavorable tumor burden, we further evaluated OS and DFS by stratifying patients according to the Milan criteria and MVI (Figure 4). Among the subgroups of patients within or beyond the bounds of the Milan criteria, preoperative GGT levels divided all patients into three groups with significantly different OS (both P < 0.001) (Figures 4(a) and 4(c)) and DFS (P < 0.001, P = 0.023, respectively) (Figures 4(b) and 4(d)). When the subgroup of patients with MVI was stratified according to preoperative GGT level, significant differences in OS were not found (P = 0.299) (Figure 4(e)), while the DFS of patients in group 2 and group 3 overlapped (P = 0.037) (Figure 4(f)); however, when the subgroup of patients without MVI was stratified according to preoperative GGT levels, the OS and DFS were significantly different (both P < 0.001) (Figures 4(g) and 4(h)). For patients with BCLC grade A, preoperative GGT levels can be used to stratify all patients into three groups with significantly different OS and DFS (both P < 0.001) (Figures 4(i) and 4(j)).

3.5. A New Scoring System Based on GGT Level

To predict the prognosis of AFP-negative HCC patients, parameters recognized as significant in the multivariate analysis were used to form the GGT-based prognostic score (GBPS) as follows: GBPS = TBIL (≤17.2 = 0; >17.2 = 1) + GGT (≤37.2 = 0; 37.2-102.8 = 1; ≥102.8 = 2) + ALBI (grade 1 = 0; grade 2 = 1) + cirrhosis (absent = 0; present = 1) + tumor number (≤1 = 0; >1 = 2) + MVI (absent = 0; present = 2). We used the X-tile program and showed that the optimal cutoff points of the GBPS score were 2 and 5. The entire cohort was then divided into three subgroups (≤2; 2-5; >5), which had significantly different OS (5-year OS rate: 89.4%, 69.6%, 35.6%, respectively; P < 0.001) and DFS (5-year DFS rate: 67.2%, 47.4%, 21.7%, respectively; P < 0.001) (Figures 5(a) and 5(b)). The GBPS model had a higher area under the curve value than the TNM stage and BCLC stage for AFP-negative HCC patients (0.726, 0.664, 0.569) (Figure 5(c)).

4. Discussion

Although serum AFP is a well-established prognostic marker in HCC, nearly one-third of HCC patients are AFP-negative. Our study showed that preoperative GGT levels can be employed as another effective marker to replace AFP as a predictor of prognosis in AFP-negative HCC patients, helping stratify the AFP-negative HCC patients who are at high risk of death and early recurrence. Moreover, the measurement of GGT is reliable and inexpensive and extensively applied in clinical practice.

Hu et al. reported that high serum levels of GGT are associated with an increased risk of liver cancer [14]. Zhang et al. and Guiu et al. found that the GGT level served as an important prognostic factor in patients with intermediate HCC treated with TACE [17, 19]. Ma et al. reported that the serum level of GGT was a convenient prognostic marker for OS and recurrence in HCC patients undergoing RFA [18]. Fu et al. reported that GGT was a promising and reliable prognostic marker in patients following liver transplantation [16]. The cutoff values of preoperative GGT in patients with HCC varied in these studies. Although molecular markers predicting HCC prognosis have been studied extensively [5, 26–28], data on the outcomes in AFP-negative HCC patients are limited.

Using the cutoff points of 37.2 U/L and 102.8 U/L determined with the X-tile program, we divided AFP-negative HCC patients into three groups with distant prognoses. In this study, we demonstrated that the preoperative level of GGT is associated with unfavorable tumor factors, including larger tumor size, multiple tumor number, and MVI, which was consistent with the findings of previous studies on gynecological cancer. Tumor size and tumor number are both independent risk factors for HCC. The prognosis of patients worsened when tumor size and tumor number increased. Defining tumor characteristics that exceed the Milan criteria that significantly impact the prognosis of HCC after resection is thus important. Additionally, previous studies showed that the presence of MVI indicated aggressive behavior of HCC and predicted a worse prognosis after liver resection [29, 30]. The early spread of cancer cells via the vasculature may be a key mechanism underlying metastasis and recurrence. However, some patients without MVI still suffer tumor recurrence shortly after curative resection in clinical practice. Thus, considering the imbalance of the baseline characteristics and to avoid being overly assertive, we further evaluated the prognostic role of the preoperative level of GGT in the subgroup of patients who met the Milan criteria and had MVI. The results showed that the preoperative GGT level was able to accurately stratify patients according to their risk level in both the subgroups of patients meeting and exceeding the Milan criteria. For patients without MVI, the preoperative GGT level may serve as a complementary marker for prognostic stratification.

As a principal enzyme involved in glutathione metabolism, GGT can exert prooxidative effects at the membrane surface and in the extracellular microenvironment. Preceding studies described GGT as contributing to persistent oxidative stress, which is a factor in genomic instability [31] and modulation of the process involved in tumor progression [9, 32]. Another theory is that GGT is induced by inflammatory cytokines, such as tumor necrosis factor-α and interferon [33], which are involved in carcinogenetic processes and the regulation of the tumor microenvironment [34, 35]. However, the mechanisms underlying the association between preoperative GGT and the prognosis of patients with HCC remain unclear and require further basic research.

Because this was a retrospective study, certain biases might exist. We enrolled a large cohort of 678 patients to address this limitation. Second, as the majority of patients had evidence of HBV infection, our data require validation in other study groups in whom HCV infection is the prevailing etiology of chronic liver disease.

In conclusion, the preoperative GGT level serves as a feasible prognostic factor for AFP-negative HCC patients following resection.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Authors’ Contributions

Liang-He Lu, Wei-Wei, and Anna Kan contributed equally to this work.