Background
Hepatocellular carcinoma (HCC) is the third leading cause of cancer death worldwide and remains a major public health concern [
1]. The established risk factors for HCC include chronic hepatitis B virus (HBV) and hepatitis C virus (HCV) infection, exposure to dietary aflatoxin, non-alcoholic steatohepatitis (NASH), alcohol-induced cirrhosis, obesity, smoking, diabetes, and iron overload [
2]. The primary prevention strategy by HBV immunization is a milestone in reducing the incidence of HCC in children [
3]. Secondary prevention by screening or surveillance in patients at a high risk of HCC is the strategy for reducing the associated mortality by providing interventions in the early stage of HCC [
4,
5].
Chemoprevention is another attractive strategy for reducing the cancer incidence by administering drugs, typically for other reasons. For example, since angiotensin II stimulates neovascularization and could act as a growth factor for cancer, angiotensin-converting enzyme inhibitors (ACEIs) and angiotensin II receptor blockers (ARBs) could conceivably reduce cancer risk [
6,
7]. Results of the study done by Chiang et al. suggested ACEI/ARB use lowers cancer risk [
8]. On the contrary, a meta-analysis of randomized controlled trials suggests ARBs are associated with a modestly increased risk of new cancer diagnosis [
9].Whether ACEIs or ARBs reduce cancer risk remains an issue to debate [
7].
Though the chemopreventive effects of ACEIs and ARBs against HCC development and recurrence have been demonstrated in animal studies [
10‐
12] and small clinical studies [
13‐
15], large-scale study remains lacking. Numerous other medications have been reported to be associated with HCC chemoprevention, including HBV medications [
16], interferons [
17], metformin [
18‐
20], statins [
21,
22], aspirin [
18], and nonsteroidal anti-inflammatory drugs [
23,
24]. Similar as ACEI and ARB, some of the results are inconsistent and remain controversial [
25,
26]. Considering that multiple risk and protective factors may coexist in patients at a high risk of HCC in real-world settings, the true comparative and competitive effectiveness of these chemopreventive drugs has not been comprehensively investigated.
The precise biological mechanisms underlying the protective effects of ACEIs or ARBs against cancer development is that the angiotensin I–VII levels increase during the inhibition of the ACE–angiotensin II–angiotensin II type 1 receptor (AT1R) axis, resulting in the activation of the Mas receptor and subsequent inhibition of cell proliferation and angiogenesis [
27,
28]. Moreover, ACEIs or ARBs have been reported to reduce liver fibrosis in human studies [
29,
30]. Further studies are warranted to determine whether these positive scientific findings can be used to extrapolate the efficacy of ACEIs or ARBs into clinical practice and translate this rationale into effective health intervention for high-risk populations [
31]. The present study analyzed the comparative effectiveness of ACEIs and ARBs in the chemoprevention of HCC in high-risk cohorts.
Methods
This was a retrospective study using nationwide cohorts of HBV and HCV patients with hypertension identified from a pseudonymied database. The primary outcome was HCC occurrence and the main exposure was the use of ACEIs or ARBs. The Institutional Review Board of National Taiwan University Hospital, Taipei, Taiwan, approved this study (NTUH REC: 201601007 W) and the need for informed consent was waived.
Data acquisition
All original data were retrieved from three linked national databases covering the entire population of Taiwan from 2005 to 2014: Taiwan’s National Health Insurance Research Database (NHIRD), the Registry for Catastrophic Illness Patient Database (RCIPD), and the Cause of Death Database. Histological confirmation or typical imaging presentation of HCC is required for patients to be registered in the RCIPD [
13].
Cohort selection
Patients who received antiviral agents that fulfilled the reimbursement criteria for HBV (HBV medications) or HCV (interferons) therapy between 2007 and 2011 were identified, with the index date defined as the date of the first prescription. The reimbursement program for HBV and HCV therapy is basically a cost-effective policy in Taiwan and is strictly audited by definite evidence of diagnosis and laboratory data [
32,
33]. Specialized case managers in each hospital guaranteed data accuracy. The enrolled population had higher risks of complications, including HCC, because of viral infection [
33].
Among the enrolled patients, those with hypertension were further selected for analyses. The diagnosis of hypertension was confirmed according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes (ICD-9-CM 401–405) in at least two outpatient visits within 120 days or at least one hospitalization and prescription of antihypertensive drugs (Anatomical Therapeutic Chemical [ATC] codes: C02, C03, C07, C08, and C09). Patients aged at least 20 years and fulfilling the aforementioned criteria from 2 years before to 6 months after the index date were included (Additional file
1: Figure S1). We excluded patients 1) diagnosed as having HCC (ICD-9-CM 155.0) before or within 6 months after the index date (to guarantee an induction period of at least 6 months after exposure to antihypertensive drugs), 2) who died within 6 months after the index date, or 3) those who received HBV medications or interferons from 2 years before to the index date.
Demographic parameters
Demographic information, namely sex, age, liver cirrhosis, and comorbidities (diabetes mellitus [DM], hyperlipidemia, malignancies other than HCC, chronic obstructive pulmonary disease [COPD], end-stage renal disease [ESRD], transplantation, and alcohol consumption), was recorded. The diagnostic criteria for liver cirrhosis and hyperlipidemia are detailed in the Additional file
2 and those for other comorbidities were described in a previous study [
34].
Chemoprevention medications
The dosage and duration of the following drugs were recorded: 1) ACEI (ATC code: C09AA), 2) ARBs (ATC codes: C09CA and C09D), 3) low-dose aspirin for antiplatelet therapy (ATC code: B01AC06), 4) metformin (ATC code: A10BBA02), 5) statins (ATC codes: C10AA, C10BA, and C10BX), 6) HBV medications, and 7) interferons (see Additional file
3: Table S1 for the comprehensive list of drugs).
Outcome measurements
The event date was the incidence of HCC confirmed by admission diagnosis (ICD-9-CM 155.0) or in the RCIPD. The patients were followed until death; withdrawal from the health insurance programs; or December 31, 2014. The date of death was obtained from the Cause of Death Database.
Statistical analyses
Demographic characteristic and exposure to medications are shown separately for the HBV and HCV cohorts. Patients were divided into two groups based on the use of ACEIs or ARBs within 6 months after the index date, to demonstrate intergroup differences (initial exposure vs. initial nonexposure). The cumulative defined daily doses (DDDs) of medications were calculated in two groups. Data are expressed as mean ± standard deviation, median (interquartile range), or number (percentage), as appropriate. The Student t test or chi-squared test was used for intergroup comparisons. The time-to-event curves of different etiological groups were plotted using the Kaplan–Meier method and compared using the log-rank test.
Cox nonproportional hazards regression models with time-dependent covariates were used to study the association of medication use with the incidence of HCC in the HBV and HCV cohorts separately. The models were adjusted for the following covariates at the baseline (fixed in time): sex, age, income, liver cirrhosis, DM, hyperlipidemia, malignancies other than HCC, COPD, ESRD, transplantation, and alcohol consumption. Moreover, the following medications were adjusted as time-varying covariates according to the actual date of treatment initiation during the follow-up: ACEIs or ARBs, aspirin, metformin, and statins. Thus, only one observation was recorded per patient, and the status values of each medication were re-evaluated at every event time for each patient whose first medication prescription date was not a missing value and was earlier than the event time.
All statistical tests were two-sided at a significance level of 0.05, and all analyses were performed using SAS Version 9.4 (SAS Institute Inc., Cary, NC, USA).
Discussion
The present nationwide cohort study of hypertensive patients with HBV and HCV infection who required antiviral therapy for alleviating the risk of primary HCC yielded four main findings. First, the estimated 7-year risks of primary HCC were 12% and 10% in the HBV and HCV cohorts, respectively. In addition, the risk did not significantly differ between patients exposed and not exposed to ACEIs or ARBs within 6 months after antiviral therapy. Second, after adjustment for comorbidities (including liver cirrhosis, DM, and hyperlipidemia) and time-dependent variables of medications (aspirin, metformin, and statins), ACEI and ARB use was not a significant predictor of primary HCC in the HBV cohort (aHR: 0.97, 95% CI: 0.81–1.16) and in the HCV cohort (aHR: 0.96, 95% CI: 0.80–1.16). Third, liver cirrhosis was a universal risk factor in both cohorts and in all subgroups. Finally, in the HCV cohort, ACEI or ARB use was associated with increased HCC risk in subgroups of patients without cirrhosis, DM, and hyperlipidemia.
The 5-year cumulative incidence of HCC in HBV patients with cirrhosis was 10%–17% [
35]; the incidence was 11% in HCV patients with cirrhosis after a median follow-up of 6 years [
36]. The present study yielded similar observations. Through regression modeling, competitive variables can be weight-based and compared. Among all comorbidities, liver cirrhosis was the strongest risk factor (aHR: 2.40, 95% CI: 1.74–3.32) in the HBV cohort, followed by DM (aHR: 1.46, 95% CI: 1.18–1.81); however, the effect of liver cirrhosis in the HCV cohort (aHR: 1.76, 95% CI: 1.23–2.50) was not this large. Because advanced cirrhosis is a relative contraindication for interferon therapy in patients with HCV [
37], the use of interferon therapy as a patient selection criterion in this study may explain the significant but relatively low effect of liver cirrhosis on HCC development in the HCV cohort. Meanwhile, HBV patients with advanced liver cirrhosis or decompensation are candidates for HBV medication in the reimbursement program of Taiwan’s National Health Insurance. Therefore, these patients were recruited in this study. Nonetheless, the comparative effect of ACEIs and ARBs in the study cohorts was marginal and insignificant. These findings are important for developing policies on the selection of chemopreventive medications against HCC in high-risk groups.
Hepatocarcinogenesis is a complex multistep process in which many signaling cascades are altered, yielding a heterogeneous molecular profile [
38]. Signaling pathways, such as the epithelial growth factor receptor (EGFR) and Ras, mammalian target of rapamycin, insulin-like growth factor receptor 1, hepatocyte growth factor and c-Met, Wingless, and angiogenesis, were, even if not entirely, involved in the complex and interactive system that formulates this hypervascular tumor entity [
38]. The phenomenon of HCC heterogeneity not only exists in different tumors [
39] but can also occur within a tumor [
40]. The contemporary view of the involvement of the renin–angiotensin system in cancer, particularly HCC, might involve EGFR transactivation by AT1R and angiotensin II-independent, antiangiogenic effects of angiotensinogen [
6,
41].The microenvironment in hepatocarcinogenesis becomes more complex in cases of viral infection and involves the transactivation of transcription factors and stimulation of inflammatory responses, thus resulting in oxidative damage, fibrosis, and genetic mutations [
42]. Renin–angiotensin system signaling may have a potential role in hepatocarcinogenesis; however, the actual real-world comparative and competitive influences associated with other signaling pathways in HBV or HCV infection remain unknown. Limited human studies have reported the potential of ACEIs in reducing HCC recurrence after curative treatment [
13‐
15]. Consistently, high gene expression of angiotensin-converting enzyme 2 in patients with HCC was associated with poor survival.
[
43]. However, no human studies have investigated the primary prevention of HCC (targeting the initiation of hepatocarcinogenesis), probably because the recurrence rate of HCC is only adequately high in patients after curative resection [
44]. Thus, the study endpoint could be reached in a manageable period.
Many possible reasons explain the gap between laboratory success and population insignificance. The dose–response relationship in vitro may not be applicable in standard dosage for hypertension. The doses recommended for cancer treatment are typically higher than the regular doses for treating non-cancer diseases. For example, the suggested dosing of everolimus, a mechanistic target of rapamycin inhibitor, in renal cell carcinoma is 10 mg per day, while it is 1.5 mg per day as an immunosuppressant in the setting of renal transplantation [
45]. Therefore, there exists a translational inconsistency between laboratory and clinical practices. Our results can save the time and effort of researchers when conducting future human investigations.
Statistical results derived from a population-based database typically yield narrow point estimates and show statistical significance with ease, even when the absolute differences between groups are too small to have biological relevance (e.g., age difference of less than 1 year in adults). Therefore, the clinical significance may warrant further confirmation. The results may sometimes be misleading and induce population panic [
46,
47]. We used multivariate Cox regression including time-dependent variables for medication use to carefully minimize the time-related bias associated with potential confounding medications in an observational study [
25]. The neutral (or considered “negative”) effectiveness of ACEIs and ARBs in high-risk cohorts suggested that their association with the prevention of HCC, if it ever existed, was weak.
In human studies supporting the protective effects of ACEI or ARB [
13‐
15], the outcome was HCC recurrence, which was different from the current study. Patients with history of HCC carry a much higher risk of HCC recurrence after treatment (and hence, a larger effect size) than those without history of HCC. Use of ACEI or ARB might, therefore, provide a “sufficient cause” for outcome prevention in the “super risky” population. Based on the results of this study, we estimated that a total of 34,058 and 19,274 event cases for HBV and HCV cohorts, respectively, are required to reach statistically significant. The effect sizes would be unrealistic since the current study already included the high-risk nationwide population at a largest and eligible scale.
In contrary, our study suggested that the use of ACEIs and ARBs was associated with an increased risk of HCC in HCV patients without cirrhosis, DM, and hyperlipidemia. ACEI exposure has been shown to be associated with breast cancer recurrence [
48]. However, the mechanism underlying this observation and whether the impact is cancer-specific remain unknown [
48]. As reported for ACEI, also ARB seem to increase new-cancer occurrence of lung [
9], breast [
49] and prostate [
49], probably due to the unopposed effect on angiogenesis through angiotensin receptor 2 stimulation under angiotensin receptor 1 blockade [
49]. Further study is warranted to confirm our finding and explore the pathophysiological mechanisms.
This study has some limitations. The NHIRD is a claims data source, which might lead to misleading findings if the study is solely based on it without validation. However, the selected cohorts (patients who received HBV and HCV therapy) in our study were strictly audited, and the outcome, HCC incidence, was retrieved from the RCIPD, a stringent certificate database. These confirmations minimized the bias of uncertainty in our study. We could not identify NASH patients accurately by ICD-9 classification, which is also a risk factor for HCC [
50]. Future studies using the 10th version of diagnosis coding, in which there is a specific code for NASH, will help resolve this limitation. The other limitation of this study is that it did not consider some dietary [
51,
52] and lifestyle-modifying factors, such as smoking [
53], which might be involved in hepatic carcinogenesis and patient survival. Direct-acting antiviral therapy, which may reduce HCC incidence after HCV eradication [
54], was not accessible in our cohorts. Furthermore, we did not investigate whether liver fibrosis can be resolved using ACEIs and ARBs.