Association of Statin Use With Risk of Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality

Key Points Question Is regular statin intake associated with a reduced rate of new liver diseases and liver-related death in the general population? Findings This cohort study, involving more than 1.7 million people after matching, found that regular use of statins was associated with a 15% lower hazard ratio of new-onset liver disease and a 28% lower hazard ratio for liver-related deaths compared with no statin use. Moreover, a decrease in hazard ratio of up to 74% was demonstrated for hepatocellular carcinoma in regular statin users compared with non–statin users. Meaning These findings suggest that statins may be a therapeutic option for liver disease prevention and should be examined in further studies.


Introduction
Worldwide, more than 2 million people die of liver-related deaths every year. 1,2In view of the increasing health crisis and limited options for liver disease prevention, drug repurposing represents an appealing prospect.Statins, as 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors, possess HMG-CoA-dependent and HMG-CoA-independent mechanisms, including antiproliferative, 3,4 antimetastatic, 5 proapoptotic, 6,7 antiangiogenic, 7,8 and immunomodulatory 9 modes of action.A previous study 10 found that statins reduce the risk of hepatocellular carcinoma (HCC) occurrence through pleiotropic effects.Furthermore, statins have been found to reduce hepatic inflammation by inhibiting the prenylation of small guanosine triphosphate hydrolases to mitigate oxidative stress. 11,12In addition, the low-density lipoprotein-lowering effect of statins may also exhibit hepatoprotective effects with regard to the development of steatosis, although further experimental studies are required. 13,14rthermore, the administration of statins can be effective in portal hypertension. 15By inhibiting RhoA translocation, statins reduce RhoA kinase activity, which counteracts further vasoconstriction. 14,16Moreover, statins improve endothelial dysfunction by enhancing endothelial nitric oxide synthesis activity and nitric oxide availability. 14,17However, information is lacking on the relevance of these effects in the general population, especially in individuals without a history of liver-related diseases.In this study, we investigated the association of statin use with both liverassociated morbidity and liver-related mortality in more than 1.7 million individuals from 3 cohorts.

Methods
This cohort study used data collected from the UK Biobank (UKB), the Penn Medicine Biobank (PMBB), and the TriNetX research network.Protocols for the PMBB study were approved by the institutional review board at Penn Medicine.Written informed consent was obtained.The UKB study has approval from the North West Multi-centre Research Ethics Committee.TriNetX received a waiver from the Western Institutional Review Board as a federated network because it only includes aggregated counts and deidentified information statistical summaries.This study conformed to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

UK Biobank
The UKB was authorized by the North West Multi-centre Research Ethics Committee and involved 502 511 volunteers aged 37 to 73 years starting in 2006.The end of follow-up was defined as death or end of data collection in May 2021.The mean (SD) follow-up time was 11.9 (1.9) years.Diagnoses were coded in accordance with the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10).The hospital admission code was used to register ICD-10 codes and transmit them to the UKB.Using the National Death Registry, we collected data on the patient's age and primary ICD-10 diagnosis that led to death.

Medication
Medication was collected at the time of entry into the study.The type and number of prescription drugs were listed by the participant.Medications taken regularly as part of a daily, weekly, or monthly

JAMA Network Open | Gastroenterology and Hepatology
Statin Use and Risk of Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality routine should be listed by the participant, so regular use can be assumed for statin users.
Medications in the UKB have been validated and analyzed before. 18We included single preparations of statins in the analyses and report the used UKB characteristics in eTable 1 in Supplement 1.

Exclusion Criteria
Criteria for exclusion from the UKB cohort were missing body mass index (BMI), age, and survival data; HIV (ICD-10 codes B20-B24); or chronic hepatitis (ICD-10 code B18) (Figure).We excluded individuals with any liver disease diagnosis (ICD-10 codes K70, alcohol-associated liver disease; K71, toxic liver disease; K72, hepatic failure, not elsewhere classified; K73, chronic hepatitis, not elsewhere classified; K74, fibrosis and cirrhosis of liver; K75, other inflammatory liver diseases; K76, other diseases of liver; and K77, liver disorders in diseases classified elsewhere) or HCC (ICD-10 code C22.0) at baseline as well as pathological alcohol consumption (>60 g/d for men and >40 g/d for women). 19Alcohol consumption and quantity were evaluated using a cascading questionnaire.First, it was determined how often alcoholic beverages were consumed.The second question was asked when a choice other than never was given and referred to the amount of alcohol consumed in a day.
A drink was evaluated by the UKB as 1 U of alcohol.The third question asked how many alcoholcontaining drinks were consumed on a typical day when drinking.Response options included 1 or 2; 3 or 4; 5 or 6; 7, 8, or 9; or 10 or more. 20

Penn Medicine Biobank
The University of Pennsylvania Health System provided the PMBB data set, which was approved by an institutional review board.Individuals gave written informed consent to access their electronic health records (EHRs) starting in 2013, with ongoing enrollment.Using medical record data, we identified diagnoses based on ICD-10 codes.The latter included the same diagnoses used in the UKB.

Medication
We calculated the duration of intake for prescribed medications from the EHR information on the prescription date, amount to dispense, and number of refills.In the absence of other information, we regarded a dispense amount of 30 as the minimum duration.We did not include any patients who had a shorter duration of intake in the analysis and only included patients with the requested or a longer duration of intake in the statin group.

Exclusion Criteria
Criteria for exclusion for the PMBB were incorrect or missing BMI, age, and survival data; HIV infection (ICD-10 codes B20-B24); and chronic hepatitis (ICD-10 code B18) (Figure).To define a cohort of patients without previously diagnosed liver disease as in the UKB, we used either the time of enrollment or, in the case of statin use, the time of initial use as the reference time.We excluded an additional 230 patients with preexisting liver disease (ICD-10 codes K70-K77).Another 6 participants were excluded because of a preexisting HCC diagnosis (Figure).Data on alcohol intake were not available.

TriNetX
The TriNetX research network is a federated multicenter research network that provides real-time access to an anonymized data set from participating health care organizations' EHRs.The Western Institutional Review Board has granted a waiver to TriNetX as a federated network.We identified all individuals aged 18 to 90 years who were newly prescribed treatment with statins and nonusers between January 2011 and December 2020.Follow-up ended in September 2022.The cohort included 784 397 individuals per group after matching.Clinical variables and death data are directly retrieved from the EHR via an integrated system of clinical records.Individuals had no diagnosis of liver disease (ICD-10 codes K70-K77) or HCC (ICD-10 code C22.0) at baseline.

Medication
Exposure to statins was lagged by 6 months to account for an adequate duration of drug use and possible latency period and to minimize reverse causality. 21First contact with a statin was defined as an index event.Prescriptions of statins before December 31, 2010, were excluded, so only new users were included.

Exclusion Criteria
Criteria for exclusion from the TriNetX cohort were missing BMI, age, and survival data; HIV infection; liver disease before December 2010; and the use of statins before the index date.Data on alcohol intake were not available.Further methods can be found in the eAppendix in Supplement 1.

Primary and Secondary Outcomes
Primary outcomes of the study were incident liver disease, HCC, and liver-related mortality.
We considered any new diagnosis of ICD-10 codes K70 to K77 after baseline as incident liver disease.

JAMA Network Open | Gastroenterology and Hepatology
Statin Use and Risk of Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality

Sensitivity Analyses Metabolic Profile
In the UKB, nuclear magnetic resonance-based metabolomic profiling was performed in a subgroup of 105 348 participants.Of the 168 metabolites provided by the UKB, we focused on 64 metabolites of the lipid profile, which could be assigned to the following groups: lipoprotein subclasses particle concentrations, cholesterol, triglycerides, fatty acids, phospholipids, and apolipoproteins.We also considered the glucose levels in more detail.We compared the metabolic profiles of statin users at baseline with nonusers after matching (eFigure 2 in Supplement 1).To compare the metabolites in the treatment and nontreatment groups, we calculated the log fold change.The Bonferroni correction was performed to avoid the type I error caused by multiple testing.A value of -log(p/64) >3.1 was considered statistically significant.

Intake Duration in PMBB
We performed sensitivity analyses to account for differences in intake duration.We excluded an intake duration of less than 30 days to assume sufficient intake period.We selected periods of 30 days or more, 90 days or more, 180 days or more, and 360 days or more as the respective minimum intake durations.

IPTW Model
We performed an additional analysis using inverse probability of treatment weighting (IPTW).We computed the IPTW weights based on the propensity scores and treatment assignment and used a survey design to account for weighting and clustering in the data.We then used Cox proportional hazards regression models to estimate the effect of statin use on each outcome of interest, including liver death, using the whole cohort without exclusions because of matching.The IPTW was used to minimize bias because of confounding when assessing the association of a treatment with the different outcomes.

Genetic Disposition
We examined the association of statin intake with liver disease outcomes in carriers of known liver disease-associated gene variants in the UKB.Genotyping was conducted in 488 377 participants using the UKB Axiom Array.Haplotype Reference Consortium and UK10K were used as a reference sample. 18,20We considered the following genes: rs738409 PNPLA3 (OMIM 609567), rs58542926 TM6SF2 (OMIM 606563), rs72613567 HSD17B13 (OMIM 612127), rs2642438 MTARC1 (OMIM 614126), and rs28929474 SERPINA1 (OMIM 107400).Because statins are predominantly metabolized via the isoenzyme CYP3A4 (OMIM 124010), we examined the association of the variant CYP3A4*22 (rs35599367), which results in partial loss of function, 22 on liver disease outcomes in the UKB in a 2:1 matching for carriers (heterozygous and homozygous) of this variant to controls. 23,24

Statistical Analysis
Data analysis was performed from April 2021 to April 2023.A Cox proportional hazards regression model was used to evaluate the association between statin use and outcomes adjusted for age, sex, BMI, ethnicity (self-reported ethnicity was included because it could affect the frequency of statin prescriptions), and number of medications (UKB only).For specific liver diseases, we performed a competing risk analysis integrating all liver diseases except the considered one as 1 competing event.
The Fine and Gray regression model with non-liver-related death as a competing event adjusted for age, sex, BMI, ethnicity, and number of medications (UKB only) was used to assess the association between statin use and liver-related death.
We verified that all variables were normally distributed.The propensity score matching approach was used to balance the characteristics of both groups at baseline (treatment group and nontreatment group) in all cohorts.The propensity scores considered sociodemographic data, health status, and medications known to affect the likelihood of statin use.Among them were age, sex, BMI,

JAMA Network Open | Gastroenterology and Hepatology
Statin Use and Risk of Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality ethnicity, diabetes (ICD-10 code E11) with or without insulin (UKB only) or biguanide intake, hypertension (ICD-10 code I10), ischemic heart disease (ICD-10 codes I20 and I25), dyslipidemia (ICD-10 code E78), and aspirin intake (and the number of medications taken, which was only available in the UKB).The UKB analyses were performed with a ratio of 5 in R software, version 4.1.2(R Foundation for Statistical Computing) (at least 1 control patient and a maximum of 5 control individuals were matched to a patient taking a statin).We also established a 1:1 matching in the PMBB and TriNetX cohorts.Additional information can be found in the eMethods in Supplement 1.
Data were expressed as mean (SD).Hazard ratios (HRs) are given with their 95% CIs.A 2-sided

Results
A total of 1 785 491 individuals were evaluated after matching.Individuals were aged 55 to 61 years on average, with a slightly higher proportion of men (up to 56% vs up to 49% women), 70% to 95% were White, whereas 2% to 3% were Asian, and 1% to 22% were Black.A total of 581 cases of liverassociated death, 472 cases of incident HCC, and 98 497 new liver diseases were registered during the follow-up period.

Association of Statin Intake With Protection From Liver Disease in the UKB
In the UKB individuals (n = 205 057) without previously diagnosed liver disease, after matching, we identified 56 109 individuals regularly receiving statins and 148 948 controls (Figure).Table 1 gives their baseline characteristics.Individuals in the UKB had a mean (SD) age of 61 (6.3) years; 44% were female and 56% were male (Table 1).Nonusers in the UKB had a mean (SD) age of 60 (6.6) years; 47% were female and 53% were male (Table 1).The proportion of individuals with diabetes was higher among statin users (14%) than nonusers (6%) (Table 1).
A total of 138 liver-related deaths, 21 new diagnoses of HCCs, and 890 new diagnoses of liver diseases were reported among statin users, whereas 406 liver-related deaths, 72 HCCs, and 2312 liver diseases were noted for controls (Figure).Statin users had a 15% (HR, 0.85; 95% CI, 0.78-0.92;P < .001)lower HR of developing incident liver disease compared with nonusers (Table 2; eFigure 1 in Supplement 1).Additional analyses of the different liver disease subdiagnoses showed a significantly lower HR for ICD-10 codes of alcohol-associated liver disease (ICD-10 code K70) and other diseases of the liver (ICD-10 code K76), which includes nonalcoholic fatty liver disease (NAFLD).

Association Between Liver Disease Risk and Duration of Statin Intake in the PMBB
To further evaluate the association with intake duration, we performed sensitivity analyses to account for differences in statin intake duration in the comparably small but ethnically diverse PMBB.
We selected a cohort of 5820 statin users and 5820 nonusers.The mean (SD) age was 60 (13) years, 45% were women, and 55% were men (Table 1).The mean (SD) age of nonusers in the PMBB was 60 ( 14) years; 46% were women, and 54% were men (Table 1).Diabetes was diagnosed in 27% of statin users and in 20% of nonusers.In the statin user group, a total of 14 liver-related deaths, 12 HCCs, c Adjustment was made to matching for the UK Biobank described in the Methods by excluding E11 from the propensity score matching for this analysis.
and 521 new liver-related diagnoses were recorded.Among nonusers, 23 liver-associated deaths, 21 HCCs, and 540 new liver diagnoses occurred (Figure).The association between statin use and liver disease protection depended on the intake duration.A short statin intake period (30-180 days)   showed no significant association, but after 360 days of statin treatment, the HR for new liver diseases was reduced by 24% (HR, 0.76; 95% CI, 0.59-0.98;P = .03)(Table 4).

Statin Bioavailability and Liver Disease in the UKB
We investigated whether the CYP3A4*22 (rs35599367) variant, which reduces the metabolism of statins, affects liver disease outcomes in people who take statins (eTable 5 in Supplement 1).We found an even stronger association in incident liver diseases among homozygous/heterozygous minor allele carriers (HR, 0.68; 95% CI, 0.51-0.89;P = .005)(eTable 5 in Supplement 1).Together, these data suggest a dose-dependent association.  a Matching was performed at a ratio of 1:1.
b Significant P value.
c Not enough end points to perform time-dependent analyses.

JAMA Network Open | Gastroenterology and Hepatology
Statin Use and Risk of Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality growth of HCC. 36Further effects on tumor development are the inhibition of MYC, 37 nuclear factor-κB cells, 38 and the protein kinase B signaling pathway, 39,40 as well as the reduced production of interleukin 6. 38 We used the UKB to establish an association of statin intake with hepatoprotection.The UKB is limited by lacking information about the length of statin intake, but it is a very well-characterized cohort with deep phenotyping as well as genetic information.This allowed us to establish a unique approach to the interplay between medication and genetics.Our findings suggest that regular statin use provided a particular benefit for PNPLA3 rs738409 minor allele carriers.However, the rare TM6SF2 rs58542926 variant, which is more likely to be harmful to the liver, 41 showed no statistically significant improvement with statins, which might be attributable to insufficient power.Similarly, SERPINA1 rs28929474 was associated with chronic liver disease, 31,32 but no statin-related beneficial association was observed for minor allele carriers.HSD17B13 rs72613567 and MTARC1 rs2642438 variant carriers, which have been associated with lower rates of NAFLD, benefitted less from statins as noncarriers. 28,42Nevertheless, it would be desirable to perform the analyses in larger cohorts, especially for homozygous minor allele carriers, to exclude that the lack of significance is attributable to small group sizes.
In addition, the UKB allowed us to study a genetic variant in CYP3A4, a key enzyme for statin metabolism, 23 which was associated with decreased expression of CYP3A4 messenger RNA. 43We demonstrated that the statin-related association on liver disease was enhanced in rs35599367 minor allele carriers.These results point toward a dose-dependent association of statins.
To counteract the limitations of the UKB, we added the PMBB and TriNetX cohorts, which had detailed information on statin intake duration.TriNetX provided the opportunity to confirm the results in a large patient cohort of more than 1.5 million individuals and to examine the association of statins with liver disease development in new users.TriNetX also enrolled individuals with more severe disease, which resulted in a higher rate of HCC cases compared with the UKB.Cross-validation of results and numerous sensitivity analyses thus had the potential to minimize misclassification bias, lead time bias, and end-of-follow-up bias.Addition of the PMBB allowed the examination of the association of statin intake duration on liver disease development in a smaller but ethnically diverse cohort.
A potential barrier to preventive treatment might be statin-associated adverse effects.
Therefore, we investigated the development of myopathy 44 but could not find any significant association for statins, which might be caused by individuals discontinuing statin use at the onset of first myopathy symptoms.Our findings confirm previous studies 45,46 that have reported an increased HR for diabetes in statin users.Remarkably, although the increased incidence of diabetes is a risk factor for chronic liver disease, it did not affect the benefits of statins with respect to the development of new liver disease or liver death in our study. 47,48is cohort study now raises the question of whether statin intake should be recommended to individuals at high risk for liver disease.Taking statins was particularly beneficial in men, individuals with diabetes, and individuals with a high FIB-4 index at baseline, which is a routine clinical score to assess a patient's risk of liver disease. 34Still, we need to confirm these associations in randomized clinical trials before recommending statins for liver disease protection.

Strengths and Limitations
A major strength of this study was the ability to confirm the results using 3 independent cohorts, which all offered longitudinal data.Nevertheless, our study had some limitations.First, the selfreported drug allocation in the UKB data set offered a potential source of error.Second, most individuals in all cohorts have not been formally screened for liver diseases.Therefore, early stages of liver disease could have remained undetected, or, in the absence of a timely diagnosis, an incorrect group allocation could have occurred.Similarly, subgroup analyses focused on individual liver diseases, and genetic analyses involved some groups with small sample sizes; therefore, the results should be considered exploratory and warrant replication in additional cohorts.Moreover, our study

JAMA Network Open | Gastroenterology and Hepatology
Statin Use and Risk of Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality is subject to the potential for type I error arising from multiple comparisons in secondary and subgroup analyses.The observational study design inherently presents the possibility of confounding, which we endeavored to mitigate through various sensitivity analyses.0][51] Moreover, the information on alcohol consumption in the UKB was based on a touchscreen interview, which may have limited reliability.
Nevertheless, self-reported alcohol consumption in the UKB has been associated with known gene loci for alcohol consumption. 52In the TriNetX and PMBB cohorts, we were unable to include any variables related to alcohol consumption.

Conclusions
In conclusion, this multinational 3-cohort study indicates a significant association between statin intake and liver disease protection, especially in men, individuals with diabetes, and individuals at (genetic) risk of liver disease.Individuals who fall into specific risk categories should undergo a thorough evaluation to determine whether they have an existing indication for taking statins because they are likely to experience significant benefits from this treatment.Finally, these findings call for clinical trials that evaluate drug repurposing for primary and secondary prevention of liver diseases.

Downloaded From: https://jamanetwork.com/ on 09/28/2023 Additional
patient characteristics were obtained from the EHR.The linkage to the EHR enabled the PMBB to receive death data and ICD-10 diagnoses leading to death.The study included 61 139 people aged 18 to 102 years, of whom 5820 were taking statins regularly(Figure).The end of follow-up was death or the end of data collection in December 2020.Mean (SD) follow-up time was 4.7 (2.3) years.

JAMA Network Open | Gastroenterology and Hepatology Initiation of New Statin Treatment and the Development of Incident Liver Disease in TriNetX
To further corroborate the results and validate our statistical approach, we developed an IPTW model to adjust for unmeasured confounding and selection bias.The IPTW method with the Cox proportional hazards regression model for competing risks showed a decrease in the HR for the development of new liver disease in statin users of 31% After establishing a clear association between statin use and liver disease protection in the UKB, we turned to the TriNetX cohort to study whether this association also holds in individuals starting a new medication with statins.The TriNetX cohort included 784 397 new users of statins and 784 397 nonusers after matching.The mean (SD) age of individuals in the TriNetX cohort was 56(11)years, P = .001)comparedwithnonusers (Table2; eFigure 1 in Supplement 1).

Table 2 .
Statin Use and the Development of Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality in the UK Biobank a Significant P value.bFor subdiagnoses, only individuals taking statins are listed, with hazard ratios and P values calculated consistently compared with individuals not taking statins.

Table 3 .
Statin Intake and the Development of Incident Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality in Individuals Without Prior Liver Disease in the TriNetX Cohort a a Matching was performed at a ratio of 1:1.b Significant P value.c

Table 4 .
Time-Dependent Intake of Statins and Risk of Incident Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality in Individuals Without Prior Liver Disease in the Penn Medicine Biobank a Cumulative Incidence of New Liver Diseases and Liver-Associated Deaths in Patients Without Prior Liver Disease in UKB eFigure 2. Metabolic Profile of Statin Users Compared to Non-users in Patients Without Prior Liver Disease in UKB eTable 1. Numerical Code of the Medication in UKB eTable 2. Statin Intake and the Development of Incident Liver Disease, Hepatocellular Carcinoma and Liver-Related Mortality in Patients Without Prior Liver Disease in UKB (Adjusted for Diet, Alcohol Intake and Socioeconomic Status) eTable 3. Analysis of the Basic Characteristics of the Cohort From eTable 2 eTable 4. Statin Intake and the Development of Incident Liver Disease, Hepatocellular Carcinoma and Liver-Related Mortality in Patients Without Prior Liver Disease in UKB in an Inverse Probability of Treatment Model JAMA Network Open | Gastroenterology and Hepatology Statin Use and Risk of Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality Influence of CYP3A4 Gene Variant and Statin Intake on Liver Health in UKB Gene Carriers Without Prior Liver Disease eTable 6. Associations of Statin Use With the Risk of Incident Liver Disease, Hepatocellular Carcinoma, and Liver-Related Mortality in Individuals Without Prior Liver Disease in Different Risk Constellations in UKB eTable 7. Overview of the Metabolites of the Volcano Plots eReferences JAMA Network Open.2023;6(6):e2320222.doi:10.1001/jamanetworkopen.2023.20222(Reprinted) June 26, 2023 15/16 Downloaded From: https://jamanetwork.com/ on 09/28/2023 eTable 5.