Association between nucleos(t)ide analog and tumor recurrence in hepatitis B virus–related hepatocellular carcinoma after radiofrequency ablation

Study Design

In this cohort study, the medical records of patients who received RFA for HCC were retrieved from the National Health Insurance Research Database (NHIRD) in Taiwan. The NHIRD contains health care data from more than 99% of the entire population of 23.38 million in Taiwan. The NHIRD comprises comprehensive information, including demographic data, dates of clinical visits, diagnostic codes, operation codes, and details of prescriptions, among others, as described in our previous studies.18, 23, 26-30 This study has been approved by the research ethics committee of the National Health Research Institutes in Taiwan.

Study Population

The patient selection process is shown in Fig. 1. We included all hospitalized patients who had received RFA as a potentially curative treatment for their newly diagnosed HCCs between July 1, 2004 and December 31, 2012. All patients who were admitted with a primary diagnosis of HCC were identified by the International Classification of Diseases, Ninth Revision (ICD-9), diagnosis code 155.0. The accuracy of HCC diagnosis was confirmed by the inclusion of patients in the Registry for the Catastrophic Illness Patient Database (RCIPD). The RCIPD is an official subsystem of the NHIRD in which histopathological confirmation or typical HCC image characteristics are required for the diagnosis of HCC.18, 23, 26, 29

Only patients with HBV infection (ICD-9 codes 070.2, 070.3, and V02.61) were included; patients with HCV infection, human immunodeficiency virus (HIV) infection, or other viral hepatitis were excluded. We also excluded patients who received combination therapies (liver resection, transarterial embolization [TAE], or intra-arterial chemotherapy) with RFA, patients with any vascular invasion (ICD-9 codes 452, 453.0, and 453.2), extrahepatic metastasis, or other malignant tumor (ICD-9 codes 140-208) before and at the first diagnosis of HCC, and patients who received antiviral therapy (lamivudine, entecavir, adefovir, telbivudine, tenofovir, or interferon) for 90 days or more before the index admission. Antiviral therapy for chronic hepatitis B has been covered by the Taiwan National Health Insurance (NHI) program since October 1, 2003, but reimbursement for antiviral therapy requires patients to fulfill certain criteria, such as elevated serum alanine aminotransferase (ALT; ≥2× upper limit of normal) and elevated HBV-DNA titers (>2,000 IU/mL).18 In the clinical practice of RFA treatment for HCC in Taiwan, patients would receive the secondary HCC treatment within 3 months after the potentially curative RFA if any viable tumor was found.31, 32 Using the definition of potentially curative RFA in the present study, that is, no tumor recurrence within 3 months post-RFA, patients with an outcome follow-up period less than 90 days were excluded.

Study Cohorts

Patients who received RFA as a potentially curative treatment for their newly diagnosed HCCs were divided into two cohorts according to their use of NAs. The treated cohort comprised patients who were treated with NAs for at least 90 days, and the untreated cohort consisted of patients who were never treated with NA therapy. Patients who received NA therapy for less than 90 days were excluded. Each patient in the treated cohort was randomly matched with 2 patients in the untreated cohort by age, sex, cirrhosis, and tumor size. To eliminate the immortal time bias, the untreated controls were also matched for the time period between the index RFA and the initiation of NA therapy for the treated patients.26, 33 Untreated controls were selected by matching the dates of index RFA with those of the treated patients (±180 days), and they were followed up from the same dates of NA therapy initiation for the treated patients.

Main Outcome Measurements

Treated and untreated patients were followed up for HCC recurrence after initiation of NA therapy and the date of matched post-RFA duration, respectively. HCC recurrence was defined as application of any treatment modality for HCC recurrence (liver resection, percutaneous ethanol injection, RFA, liver transplantation, TAE, or chemotherapy) after the date of initiation of outcome follow-up during the study period. Patients were followed up until the date of HCC recurrence, death, or December 31, 2012. Death was defined as withdrawal of the patient from coverage in the Taiwan NHI program.

Prognostic Factor Assessment

In addition to age and sex, several other important prognostic factors were evaluated. Information regarding the size of ablated HCC tumors was retrieved from the treatment database, and tumor sizes were divided into three categories, as follows: <3, 3-5, and >5 cm. Major coexisting diseases that might confound the outcome of interest were evaluated by ICD-9 codes at or before the index RFA, including liver cirrhosis (571.5), liver decompensation (789.5, 572.2, and 572.4), hypertension (401-405), diabetes mellitus (250), peptic ulcer disease (531-534), chronic obstructive pulmonary disease (COPD; 490-505, 506.4), acute coronary syndrome (410-414), cerebrovascular accident (430-436), chronic renal failure (585), and hyperlipidemia (272.0-272.2). All diseases listed in the Charlson comorbidity index were analyzed except for HIV infection, metastatic tumor, and concurrent other malignancies owing to the exclusion criteria of the present study.34

Medication information was retrieved from the pharmacy prescription database. Any antiviral therapy for chronic hepatitis B that was started during the study period was analyzed according to the design of this study. In addition, certain potentially chemopreventive medications, including nonsteroidal anti-inflammatory drugs (NSAIDs), aspirin, statins, and metformin, were also analyzed. Drug users were defined as those who used more than 1 tablet per month during the study period. Propensity scores were measured using a logistic model with age, sex, tumor size, cirrhosis, underlying comorbidities, and use of potentially chemoprotective drugs.

Statistical Analysis

Descriptive data were presented as median values (25%-75% interquartile ranges), and case numbers and percentages were used for study subjects in the treated and untreated groups. Cumulative incidences of HCC recurrence post-RFA were calculated, and calculated results were expressed as the estimated number along with the 95% confidence interval (CI). Because death from underlying comorbidities led us to apply informative censoring in calculating HCC incidence, patient mortality preceding HCC recurrence was considered to be a competing risk event. Cumulative incidences in competing risk data ratios were calculated and compared by using a modified Kaplan-Meier's method and Gray's method.35 After adjusting competing morality, differences in the full time-to-event distributions between the treated and untreated groups were compared by a modified log-rank test.23

After adjusting for age, sex, tumor size, cirrhosis, diabetes mellitus, and use of NSAIDs or aspirin, statins, and metformin, multivariate regression analyses were conducted to determine the independent prognostic factors for HCC recurrence. Hazard ratios (HRs) were determined by modified Cox's proportional hazard models for analysis of competing risk events. For the estimations of cumulative incidences under competing risk events, Cox's models were constructed by using the “cmprsk” package for R (http://cran.r-project.org/web/packages/cmprsk/index.html). All data were analyzed by SAS software (version 9.3; SAS Institute, Inc., Cary, NC).

Baseline Demographic Characteristics of the Study Subjects

We initially identified 111,325 potentially eligible patients admitted for the first time with primary diagnosis of HCC and registered in the RCIPD between July 1, 2004 and December 31, 2012 (Fig. 1). After excluding 62,503 patients who did not have HBV infection, 15 patients whose age or sex was undetermined, and 46,491 patients who received any HCC treatment other than RFA, 2,316 patients who received RFA for their newly diagnosed HBV-related HCCs were identified. We further excluded patients who had HCV, HIV, or other viral hepatitis, who received combination therapy with RFA, who were found to have vascular invasion, extrahepatic metastasis, or concurrent malignancies before or at the diagnosis of HCC, who used NA or interferon ≥90 days before RFA, who used NAs <90 days after RFA, and who had an outcome follow-up period <90 days after RFA. We identified a total of 850 patients (200 patients who used NAs for more than 90 days and 650 patients who never used NA after RFA) who received RFA as a potentially curative treatment for their newly diagnosed HCCs. Furthermore, patients in the NA-treated cohort were randomly matched 1:2 with patients in the untreated cohort by age, sex, cirrhosis, and the time period between RFA and initiation of NA therapy. Finally, a total of 133 patients were recruited in the NA-treated group and 266 in the untreated group for analysis.

Baseline demographic characteristics of study subjects in the NA-treated and untreated groups are presented in Table 1. There were no significant differences in the prognostic factors between the two groups, including age, sex, tumor size, liver cirrhosis, liver decompensation, major coexisting diseases such as diabetes mellitus, modified Charlson score, use of NSAIDs or aspirin, metformin, statins, and propensity score. The median age was around 60 years in both groups, and around 80% of patients were male. Tumor size in patients receiving RFA was generally small, and more than 70% of patients had tumors <3 cm. The median interval to start antiviral therapy after the index RFA was 0.42 (interquartile range: 0.04-1.21) years among the NA-treated patients. The median duration of NA exposure was 1.48 (interquartile range: 0.62-2.39) years during the study period, and the median duration of NA therapy was 1.24 (interquartile range: 0.38-2.15) years during the outcome follow-up period (3 months after NA initiation). However, more than 70% of patients had underlying liver cirrhosis before RFA, and around 10% of patients had liver decompensation. A proportion of patients (21.8% in the NA-treated group and 24.1% in the untreated group) had underlying diabetes mellitus, and more than half of diabetic patients (13.5% in the NA-treated group and 18.1% in the untreated group) were metformin users.

Cumulative Incidences of HCC Recurrence

Cumulative incidences of HCC recurrence in the NA-treated and untreated groups were calculated, and a total of 188 patients (47.1%) were found to have developed HCC recurrence in 2 years: 52 (39.1%) in the NA-treated group and 136 (51.1%) in the untreated group (P < 0.05). However, a high percentage of patients had underlying cirrhosis in this study; 12 (9.0%) in the NA-treated group and 25 (9.4%) in the untreated group died before HCC recurrence. After adjusting for competing mortality, the HCC recurrence rate of the NA-treated group was significantly lower than that of the untreated group (2-year recurrence rate: 41.8%; 95% CI: 32.9-50.6 vs. 54.3%; 95% CI: 48.0-60.6; modified log-rank test: P < 0.05; Supporting Fig. 1). Although the 3-year HCC recurrence rate of the NA-treated group was lower than that of the untreated group (52.7%; 95% CI: 42.7-62.8 vs. 57.6%; 95% CI: 50.9-64.2), statistical significance was not reached (P = 0.07; Fig. 2).

Cumulative Incidences of Overall Mortality

Cumulative incidences of overall mortality in the NA-treated and untreated groups are shown in Fig. 3. Although the overall mortality rate of the NA-treated group was lower than that of the untreated group (2-year mortality rate: 19.6%; 95% CI: 12.5-26.7 vs. 25.0%; 95% CI: 19.5-30.5), the difference was not statistically significant. In addition, the 3-year overall mortality rate of the NA-treated group was consistently lower than that of the untreated group (3-year mortality rate: 31.4%; 95% CI: 21.4-41.3 vs. 35.3%; 95% CI: 28.4-42.3), but the difference was still not statistically significant.

Multivariate Analysis of Prognostic Factors

The multivariate regression analysis for determining independent prognostic factors for HCC recurrence is shown in Table 2. Antiviral therapy with NAs was an independent prognostic factor that was associated with a decreased risk of HCC recurrence (HR, 0.69; 95% CI: 0.50-0.95; P = 0.02). In contrast, liver cirrhosis was an independent prognostic factor that was associated with an increased risk of HCC recurrence (HR, 1.64; 95% CI: 1.13-2.38; P < 0.01). Other potential prognostic factors, such as age, sex, tumor size ≥3 cm, diabetes mellitus, use of NSAIDs or aspirin, use of statins, or use of metformin, were not significantly associated with HCC recurrence in this study.

Multivariate Stratified Analyses in All Subgroups of Patients

Multivariate stratified analysis was performed in all subgroups of patients, including those ages < or ≥60 years, both sexes, those with tumor size < or ≥3 cm, underlying liver diabetes mellitus, and those who used NSAIDs or aspirin, statins, or metformin. Multivariate stratified analyses verified the association of NA therapy and decreased HCC recurrence in almost all patient subgroups (HR < 1.0; Fig. 4). These observations further confirmed the association between NA therapy and decreased risk of HCC recurrence in HCC patients after curative RFA. Furthermore, regarding the association between NA therapy and reduced HCC recurrence, statistical significance was reached in several subgroups, including male patients (HR, 0.62; 95% CI: 0.43-0.88), those without diabetes mellitus (HR, 0.65; 95% CI: 0.45-0.97), those who did not use NSAIDs or aspirin (HR, 0.59; 95% CI: 0.38-0.91), statins (HR, 0.70; 95% CI: 0.50-0.97), and metformin (HR, 0.63; 95% CI: 0.44-0.90).