Statin Use and Survival with Early-Stage Hepatocellular Carcinoma Free

This article is featured in Highlights of This Issue, p. 567

The poor 5-year survival probability of 15% for patients with hepatocellular carcinoma (HCC) is due, in part, to the delayed diagnosis—only 39% of patients with HCC in the United States are diagnosed at an early stage (1). Treatment modalities, such as transarterial chemoembolization (TACE), have been shown to be effective in lengthening survival (2); however, the limited efficacy in stage III or IV HCC and risk of compromising the hepatic blood supply limit its use, particularly in patients with decompensated cirrhosis (3). A number of systemic agents have been evaluated, with sorafenib showing the greatest benefit, but its use is limited to advanced HCC (4). There remains a need for an additional systemic therapeutic agent that improves the outcomes for HCC patients.

HMG-CoA reductase inhibitors, also known as statins, are commonly prescribed to control cholesterol levels. Mechanistic studies demonstrate that statin not only has cholesterol-lowering effects, but also antiproliferative effects on cancer cells, through reduced production of isoprenoid (e.g., farnesyl pyrophostate, geranyl pyrophosphate), a posttranslational modifier of the RAS protein. Statin-induced reduction in isoprenoids activates the Raf–MEK–ERK and the PI3K–AKT–mTOR pathways essential to survival of cancer cells (5) and also inhibits the replication of hepatitis C virus (HCV; ref. 6). Accordingly, statins have been shown to improve the antiviral activity of HCV polymerase and protease inhibitors in vitro (7), and are associated with an enhanced virologic response to peginterferon and ribavirin therapy in humans (8). In line with these pleotropic effects, statins have been reported to decrease the risk of HCC (9,10). A clinical trial also found a 9-month improvement in survival among advanced HCC patients treated with pravastatin (11). It remains unknown however, whether statins lengthen survival in patients with non-advanced HCC.

Not all statins are expected to have a specific local effect on the liver, as each varies in hepatoselectivity. Hydrophilic statins, such as pravastatin, accumulate in the liver via carrier-mediated transporters that are expressed in hepatocytes. As a consequence, hydrophilic statins are unable to enter cells in extrahepatic tissue that lacks the transporter. In contrast, lipophilic statins, such as simvastatin, enter cells in the liver and in the extrahepatic tissue via passive diffusion. The nondiscriminating uptake results in less accumulation in the liver (12). We hypothesized that hydrophilic statins may provide enhanced prognostic benefit to HCC patients in comparison with lipophilic statins based on their hepatoselective properties.

Our objective was to investigate whether there is a beneficial impact of statin use on overall survival (OS) in HCC patients in a population-based database of elderly patients with cancer. Given the in vitro evidence for anti-HCV activity, we also examined variability in the association of statin with HCC survival by HCV infection. Furthermore, considering the hepatoselective nature of hydrophilic statins, we investigated the impact of statin on survival by lipophilic classification.

We constructed a retrospective cohort of Medicare-insured HCC patients represented in the Surveillance, Epidemiology, and End Results (SEER) Program. The SEER Program consists of 18 regional or state cancer registries, which provide statistics on cancer incidence and survival in the United States. In 1991, the Centers for Medicaid and Medicare Services (CMS) collaborated with SEER to link data from the cancer registries and claims-based data in the Medicare insurance program to foster public health research on cancer patients (13). In 2006, CMS launched Medicare Part D, a prescription drug coverage plan and began to link the prescription drug event data to the cancer registries in 2007. We conducted analyses on patients with stage I or II primary HCC diagnosed from 2007 to 2009, who were continuously enrolled in Medicare Part D from 3 months before cancer diagnosis until death or the last day of the available claims records (December 31st, 2010). This selection allowed us to investigate the effect of statin use on HCC survival from the time of cancer diagnosis. To enable adjustment for pre-existing comorbid conditions that were indicated in inpatient and outpatient claims files, we further restricted the study population to persons who were continuously enrolled in Medicare Parts A and B from 12 months before their cancer diagnosis until death or end of follow-up.

A total of 6,825 Medicare recipients, ages 65 years and older, with a diagnosis of hepatobiliary cancer during the period 2007 through 2009 were identified in SEER records. Among these, we focused on patients with stage I or II disease, for whom there would be a sufficient follow-up period to observe a beneficial effect of statins. A total of 2,015 patients were diagnosed with stage I or II HCC (ICD-O-3 code 8170) during the follow-up period. All HCC was confirmed by histology, cytology, any unspecified microscope-based test or visual inspection. After excluding HCC cases diagnosed at autopsy, those with an unknown month/year of diagnosis, and those without continual enrollment in Medicare A, B, and D, 1,036 patients remained and made up the final analytic population for the current study (Supplementary Fig. S1). Patients were characterized by demographic factors including, age, sex, race, and neighborhood income, as well as clinical data, including grade, stage, and tumor size at diagnosis. The outcome of interest was OS since the date of cancer diagnosis to death or December 31st, 2010, the last date of available Medicare claims data.

Exposure to statins after cancer diagnosis was assessed both as a binary variable (ever use after cancer diagnosis vs. never use), and as an ordinal variable (high, moderate, low intensity vs. never use). Those taking statin at the time of or after the cancer diagnosis were considered to be statin users. Data on statin prescriptions were extracted from the prescription drug event file that details the timing, type, and dose of each prescription filled by Medicare Part D enrollees. High, moderate, and low intensity statin therapy followed the categorizations set forth by the American Heart Association and the American College of Cardiologists (14). Atorvastatin, pravastatin, and rosuvastatin were categorized as hydrophilic statins, whereas fluvastatin, lovastatin, and simvastatin were categorized as lipophilic statins.

Data on resection of the liver, liver transplantation, radiofrequency ablation, and TACE were extracted through the Medicare inpatient files using ICD9 procedure codes. Diagnoses of cardiovascular disease (i.e., myocardial infarction, stroke, peripheral artery disease), cirrhosis, dyslipidemia, diabetes, and obesity, were extracted through the ICD9 diagnostic codes in the Medicare claims files from inpatient and outpatient records. ICD9 procedures codes and Common Procedural Terminology (CPT) codes for percutaneous coronary intervention, coronary artery bypass graft or carotid revascularization procedure were also used to identify HCC patients with cardiovascular disease. All procedural and diagnostic codes are available as a supplement (Supplementary Methods and Materials).

We described the distribution of demographic, clinical, and comorbid factors by statin use after cancer diagnosis, and compared the frequencies between statin users and nonusers by χ² analysis. We also summarized the frequency and incidence rate of death per 100 person-years observed, and tested for differences in survival between variable attributes using the log-rank test. The strength of the association between statin use and survival was evaluated by Cox proportional hazards regression, using several modeling approaches: (i) a bivariable model incorporating statin as a static variable, treating those who had taken any statin after cancer diagnosis as users from the beginning of follow-up; (ii) a bivariable model incorporating statin as a time-dependent variable, in which those who were already on statin were considered users from the beginning of follow-up, whereas those who were not on statin but used statin later were considered nonusers until the first prescription after cancer diagnosis; (iii) a multivariable model with statin as a time-dependent variable, adjusting for all covariates considered above; and (iv) a multivariable model with statin as a time-dependent variable lagged by two months to remove potential bias from end-of-life decisions that influenced statin disuse. A time-dependent model was used specifically to remove bias due to immortal time, that is, the non-event period between diagnosis and statin initiation. We also examined the impact of different types of statin by simultaneously modeling atorvastatin, lovastatin, pravastatin, rosuvastatin, fluvastatin, and simvastatin as independent time-dependent variables lagged by 2 months, adjusting for all covariates. Potential difference in the hazard of death by lipophilicity of statin was investigated by simultaneously modeling hydrophilic and lipophilic statins also as independent variables. Dose response by statin intensity was examined by multivariable time-dependent analyses.

A secondary aim was to examine the statin-survival association by HCV infection status given the purported influence of statin on HCV replication, and virologic response to HCV treatment (6–8). We performed subgroup analyses by HCV infection status, and tested for effect modification by a two-way interaction term between statin and HCV using the likelihood ratio test. We also examined the following covariates as potential effect modifiers of the statin–survival association: age, grade, tumor size and stage, resection, cardiovascular disease, cirrhosis, obesity, dyslipidemia, and hepatitis B infection. This study was approved by the Institutional Review Board at Cedars-Sinai Medical Center (Los Angeles, CA). This study employed a limited dataset without direct identifiers and therefore did not require consent from patients.

Of the 1,036 patients included in the analysis, 363 were using statin either at the time of cancer diagnosis or following their cancer diagnosis. The statin users were more likely to be 75 years or older (54% vs. 43%), male (71% vs. 61%), or be diagnosed with cardiovascular disease (59% vs. 42%), obesity (14%, vs. 9%), dyslipidemia (71% vs. 54%), or diabetes (60%, vs. 48%) compared to HCC patients without a statin prescription. The statin users were less likely to have received a liver transplantation (3.9% vs. 7.7%) and to have a diagnosis of HCV infection (19% vs. 35%) compared with nonusers (Table 1).

There were 584 deaths among HCC patients with a median survival of 21 months. Those who were treated with statin had a longer median survival compared with those not treated with statin: (23.9 vs. 18.9 months; P value for log-rank test = 0.047). Table 2 describes the unadjusted rate of death by demographic and clinical characteristics of patients with HCC. Statin use, resection, radiofrequency ablation, and hepatitis B or C infection were associated with a lower rate of death, whereas being 75 years or older, black, living in a neighborhood with median income of <$50,000, having a higher tumor grade and stage at diagnosis, and hypertension were associated with a greater rate of death (Table 2).

Table 3 summarizes the relationship between statin use after cancer diagnosis and survival using several modeling approaches. In an unadjusted model incorporating statin as a non–time-dependent variable, statin use was significantly associated with a lower hazard of death [HR, 0.84; 95% confidence interval (CI), 0.70–1.00]. Incorporating statin as a time-dependent variable substantially attenuated the association toward the null (HR, 0.97; 95% CI, 0.82–1.16), suggesting that that immortal time bias had influenced the results for the first model. After adjusting for all factors listed Table 2 in the time-dependent model, statin use was not associated with survival (HR, 0.94; 95% CI, 0.78–1.15). Incorporating the time-dependent statin variable with a 2-month lag had little impact on the results (HR, 0.98; 95% CI, 0.80–1.20).

Table 4 summarizes the subgroup analyses that were conducted to explore potential effect modifiers of the statin–survival association. Statin use after cancer diagnosis was not associated with survival in any subgroup investigated. Of note, statin use was not associated with survival in HCC patients without a diagnosis of hepatitis C (HR, 0.92; 95% CI, 0.73–1.15), nor in patients with a diagnosis of hepatitis C (HR, 1.26; 95% CI, 0.82–1.93).

We performed several analyses to examine different effects of statin types and intensity on survival in HCC patients. The most commonly used statin was simvastatin (49% of statin users), followed by atorvastatin (35%), lovastatin (23%), pravastatin (13%), and rosuvastatin (8.5%). The majority of statin users were on lipophilic statins (88%) and on moderate dose (68%). We found that there was no particular statin compound, type or intensity that were significantly associated with increased survival. Of note, a trend toward lower hazard of death was observed for pravastatin users (HR, 0.82; 95% CI, 0.50–1.34), but the association was not statistically significant. Neither hydrophilic statin (HR, 0.80; 95% CI, 0.54, 1.19) nor lipophilic statin (HR, 1.03; 95% CI, 0.84–1.26) was associated with survival; and dose–response was not observed (Table 5).

In an analysis of over 1,000 elderly patients with stage I/II HCC represented in SEER-Medicare, we found that statin use after cancer diagnosis was not associated with survival. Our finding did not vary by HCV infection status, nor by any other comorbid conditions that we considered. In addition, we did not find that pravastatin, previously reported to increase survival in advanced HCC patients, was significantly associated with increased survival in this cohort. This is the first population-based investigation of the impact of statin use on OS in elderly patients with HCC and the first investigation regarding the utility of statin in non-advanced HCC patients. Our finding is therefore generalizable to patients with stage early-stage HCC in an elderly population living in the United States.

Our results stand in contrast with that of a randomized clinical trial examining the impact of pravastatin on survival among patients with advanced HCC (11). Results from this clinical trial demonstrated that pravastatin increased survival by 9 months (18 months in pravastatin vs. 9 months in placebo) and that tumor growth was slower in HCC patients treated with pravastatin. Pravastatin is a hydrophilic statin, which is taken up by hepatocytes by carrier-mediated transporters and therefore accumulates in the liver (12). Because of this hepatoselective property and the previous clinical trial finding, we hypothesized that pravastatin would be associated with greater survival in this population of elderly patient with non-advanced HCC. In our study, the HR attributable to pravastatin was 0.82, thus pointing to a potentially beneficial impact of this hydrophilic statin, but the association was not statistically significant and much weaker than determined in the clinical trial (12). Our study differs from the clinical trial of pravastatin treatment in advanced HCC patients in that this is an observational study, of elderly patients 65 years or older with stage I or II HCC. Thus, both methodologic and population differences could have led to differences in the results. Elderly patients may derive less benefit from chemotherapeutic treatment due to comorbidities that affect their life expectancy (15). It is also possible that statins have a greater effect in patients with tumors previously unexposed to statins. A majority of our statins users (86%) had used statin before cancer diagnosis, and thus the tumors may have developed resistance against statins. It was not reported whether the previous clinical trial had excluded patients with a history of statin use, but based on the younger age distribution, prior statin use was likely lower (12). HCC is known to be a highly heterogeneous cancer by etiology, molecular signature, and microenvironmental characteristics (16–19), with metastatic HCC manifesting a different molecular and immune signature from non-metastatic HCC. It is possible that statins have a stronger impact on certain molecular subtypes represented in advanced HCC, although it is yet unknown what signatures would render a tumor more responsive to statins. A potential source of bias in this observational study is that patients who are most sick would not have continued to take statins by either clinical recommendation to withhold drugs for comorbid conditions or because they would not have the energy to fill their regular prescriptions. We have tried to control for this bias in two ways: (i) by examining the patients with relatively better prognosis (i.e., stage I/II disease) and (ii) by conducting a lagged time-varying analysis in which the association with death was examined in relation to statin exposure 2 months before ascertainment of vital status. The latter method reduces bias due to end-of-life variations in prescription filling. Any residual bias uncontrolled by our methods is likely to cause an apparent favorable result for statins, which we did not find in our study. We note that the previous trial findings published in 2002 (12) have yet to be confirmed in a follow-up study. Since 2002, five trials to evaluate the effectiveness of pravastatin in the treatment of advanced HCC have been registered at www.clinicaltrials.gov (NCT01357486, NCT01903694, NCT01418729, NCT01038154, and NCT01075555). Of these, only one trial has reached completion, but study results have yet to be published (NCT01903694).

The strength of our study is that we analyzed a large sample of HCC patients, and that our study findings are generalizable to the United States elderly with HCC. Second, we had detailed prescription filling information with provision of timing, dose, and type of statin treatment. Furthermore, we were able to account for immortal time bias, which often affects pharmacoepidemiologic studies (20). The potential for immortal time bias arises when there is a non-event period of time between the start of follow-up and actual taking of the drug, during which patients could not have died, and were not exposed to the drug of interest. We found that immortal time bias had led to a suggested inverse association between statin and survival, which was corrected by treating statin as a time-dependent variable.

Because our study was observational in nature, it is possible that unmeasured confounders such as the level of liver injury, often measured by alanine transaminase (ALT), could have confounded the relationship. Abnormally high levels of alanine transaminase have historically contraindicated statin use, as statin has been associated with transient transaminitis (21). Thus, statin may not have been prescribed to HCC patients with high ALT. Another unmeasured confounder is circulating cholesterol, which may be lower on average than in the general population because of the compromised liver in HCC patients. Those with more severe liver disease would have had a lower level of cholesterol leading to less frequent statin use. Both unmeasured confounders would have biased the results toward an apparent protective effect of statin, which we did not observe. Our study was also underpowered to detect a statistically significant relationship for pravastatin.

In summary, we found that statin use after cancer diagnosis was not associated with survival in elderly patients with stages I and II HCC with or without hepatitis C. Further studies in younger patients should be conducted to determine the true effect of statin in the HCC population.

M.T. Goodman is a consultant/advisory board member for Johnson and Johnson. V. Sundaram is a consultant/advisory board member for Bristol-Myers Squibb, Valeant, Gilead, AbbVie, and Intercept. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C.Y. Jeon, M.T. Goodman, V. Sundaram

Development of methodology: C.Y. Jeon, M.T. Goodman, V. Sundaram

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.Y. Jeon

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.Y. Jeon, M.T. Goodman, G. Cook-Wiens

Writing, review, and/or revision of the manuscript: C.Y. Jeon, M.T. Goodman, G. Cook-Wiens, V. Sundaram

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.Y. Jeon

Study supervision: C.Y. Jeon

This research was supported by the Samuel Oschin Comprehensive Cancer Institute (SOCCI) at Cedars-Sinai Medical Center through the Donna and Jesse Garber Awards for Cancer Research awarded to C.Y. Jeon.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.