Introduction
Advanced hepatocellular carcinoma (HCC) is a fatal disease without curative measures, with a poor overall survival (OS) of <6 months (1). Although the management of advanced HCC has significantly changed over the last few years due to improved patient stratification and introduction of novel therapies, it remains debatable which treatment should be considered as the ‘standard therapy’ for advanced HCC cases (2). Sorafenib is currently recommended for HCC patients with Barcelona Clinic Liver Cancer (BCLC) stage C (3,4). This is based on two large-scale phase III randomized controlled trials. The Sorafenib HCC Assessment Randomized Protocol (SHARP) trial demonstrated that sorafenib, a tyrosine kinase inhibitor, was able to increase median survival from 7.9 to 10.7 months (5). In the corresponding Asia-Pacific study, the effects of sorafenib in delaying time-to-progression (TTP) and improving OS were further validated (6). However, despite the convincing data, the two studies failed to demonstrate a statistically significant benefit of sorafenib in patients with extrahepatic metastasis. No statistically significant beneficial effect was observed in patients with macroscopic vascular invasion in the Asia-Pacific study. Furthermore, the majority of subjects enrolled in those studies belonged to Child-Pugh (CP) class A (95% in the SHARP trial and 97% in the Asia-Pacific study). Further studies demonstrated that patients who received sorafenib with suboptimal liver function (CP class B) exhibited a poorer outcome compared to those with CP class A (7–10). As such, in BCLC stage C HCC patients with CP class B and/or extrahepatic metastasis, sorafenib may not be a priority choice and systemic chemotherapy remains a viable option.
Several combinations of existing chemotherapeutic agents have been used for phase II trials in advanced HCC; however, only a few were able to achieve a response rate of >20% (11). Among the regimens, the combination of 5-fluorouracil, mitoxantrone and cisplatin (FMP) consistently achieved a response rate of >20% in several studies. However, despite a significant response rate, severe side effects were reported, limiting the clinical use of FMP for advanced HCC (11,12). In order to overcome this difficulty, the identification of a reliable marker, capable of predicting therapeutic responses and, thus, preventing unnecessary side effects, is urgently required. To achieve this goal, a pilot genome-wide association study was conducted and a group of single-nucleotide polymorphism (SNP) markers in patients receiving the standard FMP regimen were identified (13). It was demonstrated that the leading marker, rs9679162, located on the intron of the UDP-N-acetyl-α-D-galactosamine:polypeptide N-acetylgalactosaminyltransferase 14 gene (GALNT14) was capable of predicting the therapeutic response, but not OS, in a previous small-scale validation study (13). To minimize the side effects, we conducted a study using a split-dose FMP protocol, based on the metronomic chemotherapy principle (14). The results revealed that the hematological toxicity was significantly reduced with the split-dose FMP regimen, without a significant alteration of the OS.
Recently, a prospective study demonstrated that an SNP on GALNT14 may predict the therapeutic response, as well as OS, in patients receiving chemotherapy with split-dose FMP (15). However, that study did not investigate the effect of other clinical parameters, including side effects, on the therapeutic response to split-dose FMP. In this study, we analyzed the effect of clinical parameters, GALNT14 genotype and side effects on TTP and OS after the first cycle of split-dose FMP regimen. Besides the GALNT14 genotype, we identified pretreatment α-fetoprotein (AFP) levels, on-treatment leukopenia and absence of vomiting as independent prognostic factors. These results highlight that, with careful selection, chemotherapy may be an optimal treatment option for a specific group of patients with advanced HCC in BCLC stage C, with CP class B, portal vein invasion and/or distant metastasis.
Patients and methods
Patients
This study was approved by the Institutional Review Board of the Chang Gang Memorial Hospital, Taoyuan, Taiwan. Between January, 2007 and December, 2012, a total of 129 patients were diagnosed with advanced HCC in BCLC stage C, with main portal vein thrombosis (PVT), or distant metastasis, or both. Of these patients, 118 received at least one course of therapy and at least one post-treatment imaging evaluation for outcome assessment. The remaining 11 patients were excluded, due to either failure to complete the first course of chemotherapy, inability to assess the outcome, or refusal to sign an informed consent. The clinical parameters recorded included gender, age, hepatitis B virus surface antigen (HBsAg), anti-hepatitis C virus antibody (anti-HCV), alcoholism, Eastern Cooperative Oncology Group (ECOG) performance status (16) ascites, CP classification, prior treatment, tumor size, PVT and distant metastasis. Biochemistry and hemogram analysis included bilirubin, alanine transaminase, albumin, creatinine, leukocyte count, neutrophil percentage, hemoglobin, platelet count, prothrombin time and AFP.
HCC was diagnosed by biopsy, aspiration cytology and/or high AFP levels (>400 ng/ml), plus two dynamic imaging studies (dynamic computer tomography and angiography).
Split-dose FMP regimen
The split-dose FMP regimen was modified from the standard regimen as previously described (14). The regimen was as follows: 5-fluorouracil was administered continuously via the intravenous route at a dose of 450 mg/m2 on days 1–5. Mitoxantrone was administered via intravenous infusion at a dose of 3 mg/m2 on day 1. Cisplatin was administered as an intravenous infusion at a dose of 40 mg/m2 over 2 h on day 1 with standard hydration. On the ninth day, the biochemical and hematological data were obtained. If the hepatic, renal and hematological data were satisfactory, 1/4 dose of mitoxantrone and cisplatin was administered on days 9 and 10, respectively. However, if the data indicated severe toxicity, the 1/4 doses were further delayed for one week, until recovery. The treatment was repeated every 4–6 weeks until a maximum of six courses. If grade 3/4 neutropenia and/or leukopenia were observed, granulocyte colony-stimulating factor was administered. Concomitant oral antiviral medication for hepatitis B was allowed. None of the patients received concomitant interferon-based therapy.
GALNT14 genotyping
Genotyping of GALNT14 was performed as previously described (15). Briefly, nuclear DNA was extracted and purified from the peripheral blood prior to treatment. The primers were as follows: forward, 5′-TCACGAGGCCAACATTCTAG-3′ and reverse, 5′-TTAGATTCTGCATGGCTCAC-3′, designed for PCR and direct sequencing of a 172-bp intronic region of GALNT14 covering rs9679162. The SNP was determined by sequencing data from both directions.
Survival and on-treatment side effects evaluation
At least one lesion was measurable prior to treatment in one dimension in all the evaluated patients. The objective tumor response was assessed by computer tomography every 4–8 weeks after the initiation of chemotherapy. OS was calculated from the date of treatment initiation to the date of death or last follow-up. TTP was calculated from the date of treatment initiation to disease progression. Based on the National Cancer Institute Common Terminology Criteria for adverse events version 3.0 (17), on-treatment side effects were evaluated at day 9 of FMP chemotherapy.
Statistical analysis
All the statistical analyses were performed with SPSS version 18.0 software (SPSS, Inc., Chicago, IL, USA). The data are presented as ratios (%) for dichotomized variables, as means ± standard deviation for continuous variables with normal distribution and as median (range) for continuous variables with a non-normal distribution. For comparisons between groups, the Chi-square or Fisher’s exact tests were used for dichotomized data, the two-sample Student’s t-test was used for continuous variables with normal distribution and the Mann-Whitney U test was used for continuous variables with a non-normal distribution. The parametric data were dichotomized into two groups, with the medians as cut-offs. Univariate and multivariate Cox proportional hazard models were used to estimate TTP and OS for clinical parameters, on-treatment side effects and genotypic variables. Those variables that were statistically significant in the univariate analysis at the level of P<0.05 were included in the multivariate analysis. Following categorization, the Kaplan-Meier method was used to estimate the survival probability between groups and the log-rank test was used to compare survival outcomes. P<0.05 was considered to indicate a statistically significant difference.
Results
Basic clinical data of patients with advanced HCC
A total of 118 patients were included. The basic clinical characteristics are listed in Table I. Of the 118 patients, 26 (22.0%) had an ECOG performance status score of ≥2. CP scoring revealed that 115 patients (97.5%) were CP class B. The etiological analysis demonstrated that 71 patients (60.2%) were HBsAg-positive, 42 (35.6%) were positive for anti-HCV, 8 (6.8%) had hepatitis B and C co-infection and 13 (11.0%) had non-B and non-C etiologies. Notably, alcoholism was observed in 37 (52.1%) of the patients with chronic hepatitis B, 14 (33.3%) of those with hepatitis C, 4 (50.0%) of those with hepatitis B and C co-infection and in 6 (46.2%) of those with non-B and non-C etiologies, suggesting prevalence of a combination etiology of viral infection and alcoholism (data not shown). The GALNT14 genotype analysis revealed that 30 patients (25.4%) had the ‘TT’ genotype. The tumor size measured by the greatest diameter was 7.8±4.17 cm. Distant metastasis was identified in 65 patients (55.1%), main PVT in 81 (68.6%) and both distant metastasis and PVT in 38 patients (32.2%), reflecting an advanced stage of HCC. A total of 63 patients (53.4%) had received previous treatment, including surgical resection, transcatheter arterial chemoembolization and radiotherapy, whereas none of the patients had previously received systemic chemotherapy or targeted drugs.
The presence of cirrhosis was determined by either computed tomography or ultrasonography, plus the presence of esophageal varices determined by endoscopy. Of the 118 patients, 97 (82.2%) were cirrhotic. Compared to the non-cirrhotic patients, cirrhotic patients exhibited a higher frequency of HCV infection (41.2 vs. 9.5%, P<0.001), ascites (57.7 vs. 19.0%, P=0.001), smaller tumor size (7.42±4.08 vs. 9.56±4.24 cm, P=0.033), lower metastasis rate (49.5 vs. 81%, P=0.004), lower leukocyte count (6.11±2.89 vs. 7.92±2.73×103/ml, P=0.01), lower percentage of neutrophils (67.44±10.89 vs. 71.45±7.05%, P=0.042) and lower platelet count (168±94 vs. 260±189×103/ml, P=0.039).
Favorable prognostic predictors for TTP and OS prior to treatment
The clinical parameters and the GALNT14 genotype accessed prior to treatment were analyzed by univariate analysis, followed by multivariate Cox proportional hazard analysis (Table II). It was observed that age (P=0.021), ECOG score (P=0.001), presence of ascites (P=0.023), GALNT14 genotype (P=0.016), tumor size (P=0.030), total bilirubin (P=0.039), percentage of neutrophils (P=0.018) and AFP levels (P<0.001) were associated with TTP. Following adjustment for the confounding factors, the multivariate analysis demonstrated that only GALNT14 ‘TT’ genotype (P=0.035) and AFP ≤2,800 ng/ml (P=0.001) were independent predictors of a favorable TTP.
The univariate analysis revealed that several factors were associated with OS, including ECOG score (P=0.001), presence of ascites (P=0.004), GALNT14 genotype (P=0.001), tumor size (P=0.004), total bilirubin (P=0.008), albumin (P=0.004), leukocyte count (P=0.028), percentage of neutrophils (P=0.001), prothrombin time (P=0.006) and AFP levels (P<0.001). On multivariate analysis, the GALNT14 genotype (P=0.006) and AFP levels (P=0.028) were identified as independent predictors of OS.
Association between outcome and on-treatment side effects
The on-treatment side effects in our study were leukopenia (55.1%), neutropenia (39.8%), anemia (85.6%), thrombocytopenia (58.5%), nausea (55.9%), vomiting (27.1%), mucositis (22.9%), diarrhea (24.6%), alopecia (2.5%), hepatoxicity (16.1%), skin rash (6.8%), fatigue (77.1%), renal insufficiency (5.9%), bleeding (12.7%) and infection (14.4%). Univariate followed by multivariate Cox proportional hazard analysis was performed to elucidate the association of on-treatment side effects with TTP and OS (Table III). The favorable factors associated with TTP were found to be leukopenia (P=0.008), absence of vomiting (P=0.009) and absence of skin rash (P=0.023). As regards OS, on-treatment leukopenia (P=0.027), absence of vomiting (P=0.013), absence of skin rash (P=0.001) and absence of renal insufficiency (P=0.030) were identified as favorable factors.
Identification of a subgroup of HCC patients most suitable for FMP therapy
To identify a subgroup of patients with advanced HCC with better TTP and OS, pretreatment AFP levels, GALNT14 genotype, on-treatment leukopenia, vomiting and skin rash were selected for further multivariate Cox proportional hazard analysis (Table IV). The favorable factors associated with TTP and OS were identified as AFP levels (both P<0.001), GALANT14 genotype (P=0.019 and 0.006, respectively), on-treatment leukopenia (P=0.007 and 0.009, respectively) and absence of vomiting (P=0.017 and 0.015, respectively). Four favorable factors were analyzed using the Kaplan-Meier survival method and the log-rank test was used to compare the TTP and OS (Fig. 1). A total of 59 patients with AFP ≤2,800 ng/ml had longer median TTP and OS (3.11 vs. 1.75 months, P<0.001; and 8.14 vs. 3.79 months, P<0.001, respectively). A total of 30 patients with the GALNT14 ‘TT’ genotype had longer median TTP and OS (3.11 vs. 2.11 months, P=0.014; and 5.75 vs. 3.93 months, P=0.001, respectively). A total of 65 patients with leukopenia had longer median TTP and OS (2.75 vs. 2.00 months, P=0.004; and 6.32 vs. 4.07 months, P=0.001, respectively). A total of 86 patients without vomiting had longer median TTP and OS (2.32 vs. 1.82 months, P=0.004; and 5.71 vs. 3.29 months, P=0.014, respectively). Finally, 9 patients (9/118; 7.6%) with all four favorable factors exhibited the longest median TTP and OS (10.64 vs. 2.07 months, P=0.002; and 25.50 vs. 4.50 months, P<0.001, respectively).
Since the GALNT14 genotype was identified as a favorable prognostic factor, we further performed Kaplan-Meier survival analysis for GALNT14 ‘TT’ genotype patients with either three of the newly identified factors, or all factors. Among the 30 patients with the GALNT14 ‘TT’ genotype, 13 with AFP levels ≤2,800 ng/ml had longer median TTP and OS (6.51 vs. 1.82 months, P=0.049; and 17.82 vs. 5.39 months, P=0.044, respectively). Of the patients with on-treatment leukopenia, 20 had a longer median OS (17.82 vs. 4.75 months, P=0.007), but not TTP (4.61 vs. 2.00 months, P=0.056). Of the patients without on-treatment vomiting, 24 had a longer median TTP (3.86 vs. 1.82 months, P=0.041), but not OS (8.14 vs. 4.14 months, P=0.070) (data not shown).
A total of 9 patients with the GALNT14 ‘TT’ genotype and three additional favorable factors exhibited the longest median TTP and OS (10.64 vs. 1.96 months, P=0.024; and 25.50 vs. 4.71 months, P=0.018, respectively) (data not shown).
Since pretreatment AFP level as a prognostic factor is easy to obtain in clinical practice, 59 patients with AFP ≤2,800 ng/ml were further analyzed with the Kaplan-Meier survival method using other prognostic factors. A total of 13 patients with the GALNT14 ‘TT’ genotype had longer median TTP and OS (6.57 vs. 2.75 months, P=0.017; and 17.82 vs. 6.50 months, P=0.005, respectively). A total of 33 patients with on-treatment leukopenia had longer median TTP and OS (4.46 vs. 2.50 months, P=0.008; and 17.82 vs. 5.28 months, P=0.001, respectively). A total of 44 patients without on-treatment vomiting had a longer median TTP (4.21 vs. 2.07 months, P=0.017), but not OS (9.64 vs. 4.61 months, P=0.088). Finally, 9 patients with AFP ≤2,800 ng/ml and three additional favorable factors exhibited the longest median TTP and OS (10.64 vs. 2.79 months, P=0.005; and 25.50 vs. 6.50 months, P=0.001, respectively) (data not shown).
Discussion
Current guidelines (3,4) recommend sorafenib as the first-line treatment for advanced HCC, based on the results of two phase III randomized controlled trials (5,6). However, not all advanced-stage HCC patients benefit from sorafenib. Those two studies failed to demonstrate statistically significant benefits in patients with extrahepatic spread or poor liver function. Furthermore, a retrospective study on HCC patients treated with soranefib demonstrated a trend towards worse OS from CP class A to B: the median OS was 6.1, 5.4 and 2.7 months in CP class A, CP class B (score 7) and CP class B patients (score 8 and 9), respectively (8). In the subset of patients with CP class B, the optimal treatment has not been clearly determined. Chemotherapy has been considered as a therapeutic option in patients with advanced HCC; however, the severe side effects and lack of prognostic predictors limit its clinical use. Our group previously demonstrated that the split-dose FMP regimen achieved a similar OS compared to the standard FMP (5.2 vs. 6.0 months, P=0.447), but was associated with a significantly lower risk of severe neutropenia (5.1 vs. 10.5%, P=0.0005) (14). A pilot genome-wide association study and a prospective confirmatory study further verified the predictive value of the GALNT14 genotype (13,15). For advanced HCC patients with CP class B receiving spit-dose FMP therapy, in those with GALNT14 ‘TT’ vs. those with the ‘non-TT’ genotype, the median TTP was 3.9 vs. 2.1 months, respectively (P<0.001) and the OS was 6.8 vs. 3.9 months, respectively (P<0.001).
In this study, we retrospectively investigated 118 patients with advanced HCC receiving split-dose FMP chemotherapy, 97.5% of whom were CP class B. Besides the GALNT14 ‘TT’ genotype, we further identified three favorable predictors of outcome, including AFP ≤2,800 ng/ml, on-treatment leukopenia and absence of vomiting after the first course of split-dose FMP. By simply using the AFP level, a group with better outcome (median TTP, 3.11 months; OS, 8.14 months) was identified. With the combination of GALNT14 ‘TT’ genotype, an even better median TTP (6.51 months) and OS (17.82 months) may be achieved. Furthermore, on-treatment side effects, including leukopenia and vomiting, may help predict the prognosis. Patients with four favorable factors exhibited the longest median TTP (10.64 months) and OS (25.5 months). Our study provided strong evidence that split-dose FMP chemotherapy may be considered as an effective treatment in patients with advanced HCC with AFP ≤2,800 ng/ml and/or the GALNT14 ‘TT’ genotype.
The serum AFP level is a useful marker used for HCC screening and diagnosis worldwide. Although AFP is not elevated in all patients with HCC, a high serum AFP has been associated with advanced tumor stage, including greater tumor size, bilobar involvement, massive or diffuse-type tumor, poorer differentiation and PVT (18,19). The serum AFP level was found to be significantly higher in HCC patients with BCLC stage D compared to stage A and B (18) and was also associated with prognosis. In addition, the AFP level has been considered as an important predictor of postoperative HCC recurrence and metastasis (20,21). In patients with advanced HCC treated with either sorafenib, transarterial chemoembolization, hepatic artery infusional chemotherapy or concurrent chemoradiotherapy, an early reduction of AFP was found to be a predictor of positive outcome (22–25). Furthermore, the early elevation of AFP was shown to be a predictor of unfavorable outcome in patients with advanced HCC treated with sorafenib (26). In addition, it was demonstrated that a low serum AFP level (≤50 ng/ml) was a favorable predictor for patients treated with intravenous 5-fluorouracil and subcutaneous recombinant interferon-α-2b (27), suggesting a predictive value of AFP in patients receiving chemotherapy. In this study, the patients with BCLC stage C HCC were divided into two groups by the median level of AFP (2,800 ng/ml). We demonstrated that patients with AFP ≤2,800 ng/ml treated with split-dose FMP regimen had a significantly better prognosis compared to those with AFP >2,800 ng/ml. This result may be explained in part by the fact that ephrin-A1 expression (an angiogenic factor) in HCC was shown to increase AFP levels and the ability of AFP to elicit the escape of HCC cells from immune surveillance (28,29). However, whether the AFP level is directly associated with chemoresistance of tumor cells remains unclear.
The severe side effects of chemotherapy, such as leukopenia/neutropenia, have been considered as one of the major obstacles for its clinical use. However, it was also demonstrated that there is an association between chemotherapy-induced myelotoxicity and patient outcome in a number of malignancies, including lung, breast, gastric, ovarian and colorectal cancer (30–38). A meta-analysis of 13 trials on various types of cancers (n=9,528) demonstrated a 31% reduction in the mortality risk for patients with a higher grade of neutropenia or leukopenia, compared to patients with lower-grade or no cytopenia (39). Our findings were consistent with those of previous studies on other malignancies, demonstrating that on-treatment leukopenia is also a prognostic factor in patients with advanced HCC treated with split-dose FMP chemotherapy. It was hypothesized that leukopenia/neutropenia, an indicator of bone marrow suppression caused by a particular dose of a chemotherapeutic agent, may also be a surrogate marker indicating that the same dose is adequate to provide an anticancer effect. Thus, lack of leukopenia or neutropenia may indicate an insufficient or absence of biological effect of chemotherapy.
Of note, the absence of vomiting was also identified as a favorable predictor in this study. Although this information may help predict outcome, the underlying mechanisms have not been elucidated. The presence of vomiting in patients with advanced HCC may be induced by extra-abdominal factors (chemotherapeutic agents, electrolyte abnormalities, central nervous system involvement) or intra-abdominal factors (gastroparesis, ileus, gastric outlet obstruction, bowel obstruction) (40). We hypothesized that on-treatment vomiting indicates poor enteral nutrition in patients with advanced HCC, which may lead to cachexia and compromise the response to antineoplastic therapy.
This study had several notable limitations, including its retrospective nature, limited sampling and restriction of the sample to a Chinese population. Our eligibility criteria confined the treatment regimen to split-dose FMP for interpretative clarity. Therefore, it may not pertain to other chemotherapeutic regimens. It is also unclear whether the leukocyte count measured at day 9 of chemotherapy actually represents the nadir in each patient. However, despite these drawbacks, our study is considered to be suitable for routine clinical practice.
In conclusion, it was recently demonstrated that ideal therapeutic results were not achieved by sorafenib in patients with advanced HCC with CP class B (9). In such patients, we identified a subgroup exhibiting a better clinical outcome when treated with the split-dose FMP chemotherapeutic regimen, simply by assessing the AFP level and/or the GALNT14 genotype. The performance may be further improved by including on-treatment leukopenia and absence of vomiting as predictors. In patients with 4 favorable prognostic factors, a median TTP of >10 months and an OS of >25 months may be expected.