Statin use and infections in Veterans with cirrhosis
Dr K. L. Lapane, Department of Quantitative Health Sciences, University of Massachusetts, 55 Lake Avenue North, ASC Building; 7th Floor, Room 7-2071, Worcester, MA, USA.
Evidence about the beneficial effects of statins on reducing infections is accumulating. Identifying ways to reduce infection risk in patients with cirrhosis is important because of increased mortality risk and costs associated with infections.
To estimate the extent to which statin use prolongs time to infection among patients with cirrhosis.
We identified Veterans with cirrhosis, but without decompensation (n = 19 379) using US Veterans Health Administration data from 2001 to 2009. New users of statins were identified and propensity matched to non-users and users of other cholesterol-lowering medications (1:1 matching). The cohort was followed up for hospitalisations with infections. Cox regression models with time-varying exposures provided estimates of adjusted hazard ratios (HR) and 95% confidence intervals (CI).
New statin use was present among 13% of VA patients with cirrhosis without decompensation. Overall, 12.4% of patients developed a serious infection, and 0.1% of patients died. In the propensity-matched sample, statin users experienced hospitalisations with infections at a rate 0.67 less than non-users (95% Confidence Interval: 0.47–0.95).
Infections are a major concern among cirrhotic patients and have the potential to seriously impact both life expectancy and quality of life. Statin use may potentially reduce the risk of infections among patients with cirrhosis.
Over 27 000 deaths from cirrhosis and its complications occur annually. In patients with cirrhosis, infections increase mortality and drive costs of care higher.[2-4] Cirrhotic patients with infections may experience death due to infection, or an acceleration of disease progression from hepatic encephalopathy and renal failure, rendering them ineligible for liver transplantation.
The use of statins, HMG-CoA reductase inhibitor agents, has recently emerged as a protective factor for infections. Statins may reduce the risk of cancer,[6, 7] cataracts, Alzheimer's disease and mortality from sepsis and other infections. While the biological mechanism explaining these findings remains unknown, the anti-inflammatory and immunomodulatory properties of statins are thought to be involved,[11, 12] and in the immune system, statins may increase the ability of phagocytes to create extracellular traps.
Given biological plausibility, coupled with emerging research supporting the hypothesis that statins reduce infections, we decided to explore this potential relationship in a cohort of cirrhotic veterans. On the basis of previous reports, we hypothesised that statin users would develop serious infections at lower rates compared with non-users. The large data source available from the United States Veterans Health Administration's national shared electronic medical record system allowed us to compare statin users with non-users, and other cholesterol-lowering medication users. The primary aim of this study was to determine if negative outcomes (i.e. death or infections coded during a hospital stay) were present at lower rates in patients initiating statin medications compared with non-users of statins among veterans with early cirrhosis.
The McGuire VA Medical Center Institutional Review Board approved this protocol.
Using data from the Veterans Health Administration's (VHA) Austin Information Technology Center (AITC) database from 2001 to 2009, we conducted a retrospective cohort study. The VA Health System provides care to over 5 000 000 veterans broadly representative of the US veteran population in terms of sex, race, age and geography. The VHA has a centralised database containing patient records from all nationwide VHA centres, which we linked to the VHA's Pharmacy Benefits Management (PBM) database containing medication information. The VA Decision Support System and out-patient medical care datasets prospectively collect demographic and other clinical characteristics, eliminating the possibility of evaluation of liver disease severity using Meld and Child Score.
We identified 123 036 patients with a primary or secondary diagnosis of cirrhosis (ICD-9-CM codes 571.2, 571.5, 571.6). These ICD-9 codes have been used in past studies to identify patients with cirrhosis.[17, 18] Patients were considered diagnosed with cirrhosis if they had ≥2 out-patient visits between 2001 and 2009 (n = 94 704). Patients also had (i) evidence of VA Pharmacy activity i.e. ≥1 prescription (n = 86 191) and (ii) ≥1 year of follow-up data available and a 6-month washout period for statins (n = 35 280). Patients were excluded if taking the drug rifaximin, which may indicate hepatic encephalopathy, had infections during the washout period, or were on both statins and other cholesterol-lowering drugs (n = 29 810). We intended to analyse patients with decompensation separately owing to the increased risk of infections in this group, but the small sample size precluded this. These patients had one or more of the following codes signifying decompensation: (hepatic encephalopathy (572.2); hepatorenal syndrome (572.4); spontaneous bacterial peritonitis (SBP) (567.23); ascites (789.5); variceal bleeding (456.0, 456.2). The cohort included 19 379 patients.
The primary outcome was time to first serious infection documented during an in-patient hospitalisation or death. In-patient infections specifically extracted included (a) skin infection, including cellulitis 680–686; (b) lower respiratory infections 480–487; (c) SBP/peritonitis: 567.x; (d) urinary and kidney infections (UTIs) 590.x; (e) Clostridium difficile 008.45; (f) infectious gastroenteritis 0090; (g) systemic inflammatory response syndrome (SIRS) 995.91; and 995.92 (h) sepsis and bacteraemia (in the absence of a primary infection) including the following: (i) bacteraemia 790.7 and (ii) septicaemia 038.x. Previous work has confirmed the accuracy of VA-coded infections. The rank order of types of infections observed was lower respiratory infections, skin infections, sepsis, bacteraemia and Clostridium difficile infections.
The VA Pharmacy Benefits Management database was used to determine statin and other medication use, and includes details of out-patient medication use for all veterans, including name of medication, number of doses, fill dates, quantity dispensed and supply days. Exposure was defined dichotomously as receipt or no receipt of statins during each 90-day window from index date for each patient. Statins included atorvastatin, cerivastatin, fluvastatin, lovastatin, pravastatin, rosuvastatin and simvastatin. We required users to have received at least a 90-day supply. We also considered other cholesterol-lowering medications including niacin, fibrates, resins and selective cholesterol absorption inhibitors. Non-use was defined as absence of a prescription for a statin or other cholesterol-lowering medication.
Confounders included alcoholic aetiology of cirrhosis, comorbid conditions and concurrent medications such as proton pump inhibitors (PPIs) and nonselective beta blockers.[23, 24] Age, race, year of entry into the cohort, Hepatitis C virus status, H2 receptor antagonist use, number of in-patient and out-patient visits and the Charlson comorbidity index were considered as well. The Charlson comorbidity index was based on data in the first year after cirrhosis diagnosis. Using ICD-9 codes in the VA out-patient and in-patient treatment files, the diagnostic conditions included were myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatological disease, peptic ulcer disease, cirrhosis, hepatic failure, diabetes mellitus, diabetes with chronic complications, hemiplegia or paraplegia, chronic renal disease, malignant neoplasms, myeloma or leukaemia, lymphomas, metastatic solid tumours and HIV/AIDS.
Propensity score analyses[26, 27] were performed. With this technique, a sample of patients who are similar on all potential confounders except statin use is created. Two matched propensity samples were created: (i) statin users matched to other cholesterol-lowering medication users; and (ii) statin users matched to non-users of any cholesterol-lowering medications. The propensity score is a summary variable created from a logistic regression model including all risk factors for serious infections as covariates. The C statistic suggested good overlap of the propensity score distributions for statin users and non-users (C statistic = 0.6). The new statin users were 1:1 matched to patients who initiated other cholesterol-lowering medications using a method that improves inexact matching by making best matches first, then taking the next best matches in a hierarchical sequence. We were able to match 503 statin users to other cholesterol medication users and 1760 statin users to non-users.
A visual inspection of the balance of confounders across statin use confirmed the adequacy of the procedure. Person time was calculated as time from study entry (new use of statins, new use of other cholesterol-lowering medication or adjusted index date for matched non-user) to first serious infection, loss to follow-up, death or end of study. Separate Cox regression models were developed to estimate the association between statin use vs. non-use and statin use vs. other cholesterol-lowering medication use, adjusting for confounding and considering statin use and other cholesterol medication use as time-varying covariates. The models produced hazard rate ratios and 95% confidence intervals. Previous work demonstrates the importance of considering time-varying drug exposures, as otherwise the reliance on time-fixed methods may introduce bias.
The median time of follow-up was 1194 days, with a minimum of 365 days of follow-up and a maximum of 3103 days of follow-up. Thirteen per cent initiated statin use. The majority (49.3%) of nonstatin cholesterol-lowering drug users were under age 55, which was not true for either the statin users or the non-users of any cholesterol-lowering medications (Table 1). Fewer users of nonstatin cholesterol-lowering drugs (14.3%) had a Charlson comorbidity score of 5 or above compared with statin users (17.1%) or non-users of any cholesterol-lowering drugs (16.0%). Thirty-six percent of statin users had alcohol-related cirrhosis compared to 27.2% of nonstatin cholesterol-lowering drug users and 33.1% of users taking no cholesterol-lowering drugs. Fewer statin users (39.6%) had Hepatitis C virus compared with nonstatin cholesterol-lowering drug users (52.1%) and non-users of any cholesterol-impacting drug (57.9%).
Table 1. Comparison of characteristics between statin users, other cholesterol-lowering medication users, and non-users among veterans with cirrhosis without decompensation
|Drug use at baseline|
|Proton pump inhibitor||23.5||25.7||24.6||0.4393|
|Nonselective beta blockers||5.4||5.6||6.2||0.2861|
|Out-patient visits in past year (median, IQR)||8 (16)||10 (17)||9 (16)|| |
Table 2 shows the demographic characteristics within the propensity-matched groups. Age and race were relatively even among groups, with white and unknown race remaining the most predominant. Drug use was also relatively even among groups, with proton pump inhibitor use present in one-fourth of patients and nonselective beta blocker use in 5% of patients. Alcohol-related cirrhosis accounted for approximately 30% of cases among the group of statin users matched to nonstatin cholesterol-lowering drug users, and approximately 35% of cases among the matched group of statin users and users of no cholesterol-lowering drugs. Hepatitis C virus was present among approximately 52% of patients in the matched group of statin users and nonstatin cholesterol-lowering drug users, and approximately 40% of patients in the matched group of statin users and non-users of any cholesterol-lowering drugs.
Table 2. Characteristics of propensity-matched samples of statin users, other cholesterol medication users, and non-users among veterans with cirrhosis without decompensation
|55–64||33.2||34.4|| ||41.1||36.0|| |
|65–74||11.1||13.3|| ||17.2||15.7|| |
|75+||4.0||3.0|| ||6.0||7.1|| |
|Black||10.9||11.1|| ||8.5||10.2|| |
|Hispanic||3.8||3.6|| ||5.5||6.1|| |
|Other||1.2||0.8|| ||0.3||0.5|| |
|Unknown||48.9||46.5|| ||48.0||44.4|| |
|Drug use at baseline|
|Proton pump inhibitor||29.6||25.7||0.1585||24.5||22.3||0.1304|
|Nonselective beta blockers||6.0||5.6||0.7868||5.3||4.7||0.3962|
|Out-patient Visits in Past Year (median, IQR)||9 (18)||10 (17)|| ||8 (16)||8 (16)|| |
|3–4||23.5||20.3|| ||26.4||23.2|| |
|5+||11.7||14.3|| ||18.4||14.6|| |
Table 3 shows statin and other cholesterol-lowering drug use, dosages, and duration of dose in cirrhotic patients. Among these patients, 2468 were on a statin only. Among statin users, 90.6% were on simvastatin, 9.4% were on lovastatin, 7.9% were on pravastatin, 3.4% were on fluvastatin, 2.4% were on rosuvastatin and 0.9% were on atorvastatin. There were 503 patients using other cholesterol-lowering medications only, including 14.5% on selective cholesterol absorption inhibitors, 36.0% on niacin and 57.1% on fibrates.
Table 3. Use of statins during the study period
|Other cholesterol-lowering drugs|
|Fibrates||1793||299|| || || || || |
During the follow-up, 12.4% (n = 2393) of patients developed serious infections a median of 607.5 days after entering the cohort, 0.1% (n = 26) of patients died and 0.5% (n = 91) of patients underwent liver transplant. The infections experienced were SBP/peritonitis (n = 73, 3.1%), lower respiratory, e.g. pneumonia (n = 946, 40.1%), skin (n = 866, 36.7%), bacteraemia (n = 171, 7.2%), septicaemia (n = 143, 6.1%), Clostridium difficile (n = 131, 5.6%) and UTI (n = 32, 1.4%). The median time to infection ranged from 394 days for kidney/urinary tract infections to 1057 days for systemic inflammatory response syndrome.
For the total sample (Table 4), both statin use and nonstatin cholesterol-lowering drug use was observed to have a protective effect (adjusted hazard ratio (aHR): 0.42, 95% CI: 0.36–0.48; aHR: 0.44, 95% CI: 0.33–0.61 respectively) as compared with non-users of any cholesterol-lowering drug. In the propensity-matched samples, statins, compared with other cholesterol-lowering medications, did not retain a protective effect; however, the model did trend towards significance. Compared with non-users, the rate of infection or death was significantly lower among statin users (aHR: 0.67: 0.47–0.95).
Table 4. Effect of statins on infections among total sample and propensity-matched samples
|Statin users||200||752||266||0.42 (0.36–0.48)|
|Other cholesterol-lowering users||40||153||261||0.44 (0.33–0.61)|
| Non-users (ref)b||2153||4516||477||1.00|
|Statin users (n = 503)||31||64||484||0.77 (0.31–1.94)|
|Other cholesterol-lowering users||40||73||551||1.00|
|Statin users (n = 1760)||154||312||493||0.67 (0.47–0.95)|
Our findings are consistent with growing evidence suggesting that statins' benefits extend beyond cholesterol lowering. Our study suggests that statins may reduce the risk of infections among some cirrhotics at particular risk of infection. The magnitude of this reduction in infections is similar to that in other studies.[13, 29]
Statin use is relatively common among veterans, and is generally considered safe among those without decompensated disease. Nearly 13% of veterans with compensated cirrhosis used statins. Previously, 20% of in-patients were found to be continuing or new statin users compared to the 37.5% of US adults with high LDL cholesterol levels. Statin use in our study may be lower partially because only new prescriptions of statin medications were considered and only statins received through the VA were considered. Most statins are available as generics; cost or convenience advantages may exist for veterans to obtain statins outside the VA system.
We found that statins provided a protective effect against infections in veterans with cirrhosis compared with non-users of any cholesterol-lowering medications. Statin use has been associated with a decreased risk of pneumonia and other severe infections including sepsis. Yet, statins may not reduce additional infections in patients in whom infections are already present. Among studies that found no role for statins reducing risk of infection, underlying health differences between statin users and nonstatin users were cited as a potential explanation. Our study supports that statins may have a role in preventing infections in certain populations. The proposed mechanism is through the anti-inflammatory properties of statins. We cannot rule out confounding by indication as an alternative explanation for these findings. The findings regarding statin use vs. other cholesterol-lowering medications were unequivocal. An alternative explanation for our findings is based on the lack of difference in infection rates between statin users and users of other cholesterol-lowering drugs. This finding may indicate that some other factors associated with patients being treated for elevated cholesterol actually lead to lower infections.
Our study benefits from sound methodology. Previous studies have not considered or controlled for confounding by indication, which we have addressed by matching statin users to nonstatin users who are taking another cholesterol-lowering agent. Our study was also able to consider the length of time statins were used. The VA data source also contributes to the strength of this study as it is largely representative of many different types of VA patients, and contains a depth and breadth of information not always possible in other administrative sources. This database has proven very useful to analyze the individual effects of several commonly prescribed drugs such as non-selective beta-blockers and proton pump inhibitors and has brought into focus the individual drugs' effects on the rate of serious infections in Veterans with cirrhosis.
This study also has potential weaknesses. Our population is older and predominantly male; our findings cannot be applied to younger patients and women. This study was unable to control for certain important confounders, such as severity of liver disease, usually measured through a MELD or CHILD score, as these measures were not available in claims data. If these factors are associated with use of statins, this could explain our findings. The matching design may have minimised the effect of potential confounders on the study findings. It is also worth noting that the biological means through which statins may reduce infection risk is not well understood. We were unable to analyse the effects in patients with decompensated cirrhosis due to insufficient sample size. Decompensated patients have higher infection rates and warrant investigation.
Despite the limitations owing to non-experimental studies including residual confounding, some have suggested a potential beneficial regimen of generic, low-cost drugs, including statins, which patients with cirrhosis could take to prolong life and improve quality of life. Such recommendations are likely premature as contradictory evidence exists.[37, 38] Indeed, some consider decompensated cirrhosis a contraindication for statin use. Yet, statin use could lower portal pressure, a risk factor for further liver complications, thereby improving outcomes among cirrhotics. Statins have been implicated as reducing the risk for other liver conditions, including a reduction in risk for hepatocellular carcinoma and a beneficial effect on liver enzymes among patients with primary biliary cirrhosis.[41, 42] A beneficial effect on liver enzymes on veterans with hepatitis C virus has been shown.
The number of patients with cirrhosis is increasing, resulting from hepatitis C, alcohol abuse and metabolic syndrome, among other causes. Liver transplantation is the only cure, and preventing worsening of cirrhosis and associated outcomes, such as infections, is the primary goal in treatment. Infections account for a large proportion of deaths in patients with cirrhosis and are an area to target to reduce morbidity and mortality. With the infection-reducing potential of statins, these drugs could play an important role in reducing mortality from cirrhosis. While their benefits are not confirmed, conducting further research is warranted.
Guarantor of the article: KLL.
Author contributions: KLL takes responsibility for the work as a whole. CMF and AP contributed to data interpretation, manuscript drafting and revision. SMR contributed to data analysis, interpretation and manuscript revision. JB contributed to data acquisition, data interpretation and critical revision of the manuscript. KLL contributed to study design, data analysis and interpretation, and manuscript drafting and revision. All authors have approved the final version of the manuscript.
Declaration of personal interests: KLL has consulted for OrthoMcNiell Janssen Scientific Affairs on pain management in nursing homes. The other authors declare that they have no competing interests.
Declaration of funding interests: ALP is supported by grant number K08HS018578 from the Agency for Healthcare Research and Quality.