Statin Use and Risk of Gallstone Disease Followed by Cholecystectomy FREE

Approximately 10% to 20% of white adults in developed countries have gallstones, and gallstone disease is a leading cause of gastrointestinal morbidity and inpatient admission in western countries.^1- 2 Colicky pain and complications such as cholecystitis, choledocholithiasis, or pancreatitis occur in a substantial portion of gallstone carriers.^3- 4 Gallstone disease represents a serious burden for health care systems worldwide; more than 700 000 cholecystectomies are performed annually in the United States.⁵

Gallstones are classified as either cholesterol (80%-90%) or pigment stones (10%-20%). Cholesterol stones are formed on the basis of cholesterol-supersaturated bile,^3,5 whereas pigment stones consist mainly of polymerized calcium bilirubinate.³ Risk factors such as age, female sex, obesity, high-carbohydrate and high-fat diet, estrogen-containing contraceptive use, and postmenopausal estrogen therapy predispose to cholesterol gallstone formation.^1,3- 5

Statins decrease hepatic cholesterol biosynthesis and may thereby decrease biliary cholesterol secretion,⁶ consequently leading to diminished cholesterol concentration in bile. Studies in animals have reported beneficial effects of statins in preventing gallstone formation.^7- 8 In humans, some authors reported diminished gallstone formation, decreased cholesterol content in bile, or gallstone dissolution,^9- 12 whereas others did not find evidence for such an effect.^13- 15 The authors of a recent observational study reported a slightly decreased frequency of cholecystectomy in women with self-reported long-term statin therapy.¹⁶ Two small studies reported conflicting results; one found no association between statin use and gallstone risk,¹⁷ and the other reported a risk reduction with limited statistical power.¹⁸

Fibrates affect lipid metabolism mainly by acting as synthetic ligands of the peroxisome proliferator–activated receptor α. Activation of peroxisome proliferator–activated receptor α affects numerous metabolic pathways including glucose and lipid metabolism, reducing level of plasma triglycerides and glucose concentration, and increasing high-density lipoprotein cholesterol levels. Studies in humans demonstrated that short-term fibrate treatment is associated with decreased bile acid synthesis and biliary excretion.^19- 20 In a randomized controlled trial, a higher incidence of cholecystectomy was noted in patients taking fibrates,²¹ and another study reported a higher prevalence of gallstones in patients taking fibrates.¹⁷

Therefore, currently available clinical studies on the possible effect of statins on gallstone formation and the risk of cholecystectomy are limited by either small sample size, lack of long-term follow-up, sex restriction, or methodological drawbacks. We conducted a large long-term observational study to explore the association between statin use and the risk of developing an incident diagnosis of gallstone disease followed by cholecystectomy.

A case-control analysis was conducted using the UK-based General Practice Research Database (GPRD). The UK-based GPRD was established around 1987 and encompasses data on approximately 5 million patients who are registered with selected general practitioners. The general practitioners provide anonymized medical information for research purposes, which are recorded in a standard manner. Patients enrolled in the GPRD are representative of the United Kingdom with regard to age, sex, geographic distribution, and annual turnover rate. The information recorded includes patient demographics and characteristics (eg, age, sex, height, weight, smoking status), symptoms, medical diagnoses, referrals to consultants, and hospitalizations. Because physicians generate drug prescriptions directly with the computer using a coded drug dictionary, all recorded prescriptions include the name of the preparation, route of administration, dose of a single unit, number of units prescribed, and intake regimen (in most instances). The database has been described in detail elsewhere,^22- 25 has been validated extensively,^26- 28 and has been the source for numerous epidemiological studies published in peer-reviewed journals.

The study protocol was approved by the Independent Scientific Advisory Committee for Medicines and Healthcare Products Regulatory Agency Database for Research. Informed consent by patients was not needed for this database study.

Patients aged 20 years or older with a first-time diagnosis of gallstone disease or cholecystectomy between 1994 and 2008 were identified based on the Oxford Medical Information System and Read codes. All patients were required to have both a recorded diagnosis for gallstone disease or a complication thereof (such as bile duct obstruction, cholecystitis, or cholangitis) and a record of cholecystectomy within 2 years of the diagnosis or a cholecystectomy only. The date of the first-time diagnosis of gallstone disease or the date of cholecystectomy in patients without a recording of a gallstone diagnosis is referred to as the index date. Patients with less than 3 years of active history in the database prior to the index date, as well as those with a history of alcohol or drug abuse, cancer (except non–melanoma skin cancer), or human immunodeficiency virus prior to the index date were excluded.

A total of 300 potential patient profiles (selected at random) were reviewed to verify the validity of the patient selection criteria.

From the base population, 4 control patients were identified at random for each patient with cholecystectomy, matched on calendar time (same index date), age (same year of birth), sex, general practice, and number of years of active history in the GPRD prior to the index date. Controls had no recording of gallstones, clinical complications suggestive of gallstones, or cholecystectomy. The same exclusion criteria were applied to patients and controls.

From the computer record, exposure was assessed to statins, fibrates, or other lipid-lowering agents (anion-exchanger resin, probucol, acipimox, niacin, fish oil, or omega fatty acids) prior to the index date for patients and controls. Patients were classified as currently taking medications if the last prescription was recorded less than 90 days, or as formerly taking medications if the last prescription was recorded 90 or more days prior to the index date. Medication use was classified by duration of use prior to the index date based on the number of prescriptions (statins: short-term, 1-4; medium-term, 5-19; or long-term, ≥20; fibrates and other lipid-lowering drugs: short- to medium-term, 1-9; long-term, ≥10). Statin use was also classified according to the tablet dose. Finally, duration and timing of use were combined into 1 exposure variable.

Conditional logistic regression analyses were conducted using SAS statistical software version 9.1 (SAS Institute Inc, Cary, North Carolina) to calculate relative risk estimates as odds ratios (ORs) with 95% confidence intervals (CIs) at a 2-sided P value of .05. With an assumed prevalence of exposure to statins of about 1%, the required sample size was about 19 000 patients (and 4 times as many controls) to detect an OR of 0.75 at the significance level of .05 with a power of greater than 90%.

The potential confounders age, sex, general practice, calendar time, and years of recorded history in the database were controlled for by matching and the ORs were further adjusted for smoking status (none, current, former, or unknown) and body mass index (BMI, which was calculated as weight in kilograms divided by height in meters squared) (BMI categories: <18.5, 18.5-24.9, 25.0-29.9, 30-34.9, ≥35, or unknown) in the multivariate model. The risk estimates were adjusted for a history of ischemic heart disease, ischemic stroke, transient ischemic attack, and use of opposed or unopposed estrogens (current or past use [defined as last prescription ≤180 days before the index date or thereafter] of 1-9 or ≥10 prescriptions). Further testing included whether other diseases (eg, liver cirrhosis, renal diseases, gastrointestinal tract ulcer, constipation, urinary dysfunction, hypertension, hypothyroidism, inflammatory bowel disease, or diabetes mellitus) or other drugs (eg, thiazides, contraceptives) confounded the main association of interest. Because this was not the case, these parameters were not included in the final model.

Various sensitivity analyses were conducted. In the first model, only patients with abdominal pain recorded within 90 days before the index date were included to reduce the likelihood of detection bias (ie, including patients who had gallstones detected by chance). In the second model, patients with cholecystectomy in the absence of a recorded gallstone diagnosis were excluded. In the third model, the index dates were shifted for all patients and their controls by 30, 90, and 180 days backward in time because gallstones may have been present prior to the actual first-time recording of the diagnosis; this was done to rule out potential protopathic bias (ie, the possibility that general practitioners may have stopped statin treatment in patients due to abdominal pain). In the fourth model, an analysis was restricted to patients and controls whose index date was before 2004 because 10 mg of simvastatin became available over-the-counter in August 2004.²⁹ In the fifth model, current long-term statin use was compared with current short-term statin use to explore the association of interest in a population diagnosed as having hypercholesterolemia; this analysis likely excluded patients with normal cholesterol levels who may have a different gallstone risk than patients with hypercholesterolemia. The final sensitivity analysis was conducted taking another proxy for duration of statin use—the total number of statin tablets prescribed prior to the index date, which equals the total number of exposure days. This analysis was performed to estimate more precisely how much exposure time it takes to lower the risk of developing gallstones followed by cholecystectomy, and because many, but not all statin prescriptions, are issued for 28 days.

A total of 27 035 patients (n = 9602 cholecystectomy only; n = 17 433 cholecystectomy after a gallstone diagnosis or associated complications) and 106 531 matched controls were identified (Table 1). The study population was 76% women and the mean (SD) age was 53.4 (15.0) years at the index date. The OR of developing an incident gallstone diagnosis with cholecystectomy substantially increased with increasing BMI.

In the study population, 11 264 (2396 patients and 8868 controls) were taking statins, 1514 were taking fibrates, and 1038 were taking other lipid-lowering agents. The majority of patients (87%) treated with lipid-lowering agents were taking statins only (ie, not in combination [concurrently or subsequently] with other lipid-lowering agents). Compared with the reference group of patients who did not take statins, the adjusted OR (AOR) of developing gallstones with cholecystectomy was 0.78 (95% CI, 0.73-0.83) for current statin use and 1.19 (95% CI, 1.07-1.32) for past statin use (adjusted in the multivariate analysis for age, sex, general practice, calendar time, BMI, smoking history, a history of ischemic heart disease, stroke, transient ischemic attack, and use of estrogens, fibrates, and other lipid-lowering agents). Compared with nonuse of statins, the AOR for 1 to 4 current prescriptions was 1.10 (95% CI, 0.95-1.27), 0.85 (95% CI, 0.77-0.93) for 5 to 19 current prescriptions, and 0.64 (95% CI, 0.59-0.70) for 20 or more current prescriptions regardless of timing of use (Table 2). The findings were similar for men and women and for those younger than 60 years or for those aged 60 years or older. The analysis also was stratified by individual statin, with similar findings for all statins, with little statistical power for some of the newer compounds (Table 3). The AORs for current statin use of 20 or more prescriptions, stratified by dose, also were assessed and the results are displayed in Table 3. Current long-term use of fibrates was associated with a slightly increased AOR of 1.39 (95% CI, 1.12-1.72) for developing gallstones followed by cholecystectomy.

Because a comparison between patients with statin use and nonuse not only compares exposure to statins, but also patients with hypercholesterolemia to patients without hypercholesterolemia, the analysis was restricted to patients with statin use (ie, patients with hypercholesterolemia). Among these patients, long-term statin use was compared with short-term statin use, yielding an AOR of 0.58 (95% CI, 0.50-0.68). When this analysis was further stratified by BMI categories, the AOR associated with long-term statin use compared with short-term use was 0.58 (95% CI, 0.41-0.81) for normal weight (BMI, 20-24.9), 0.63 (95% CI, 0.48-0.83) for overweight (BMI, 25-29.9), and 0.65 (95% CI, 0.50-0.86) for obese patients (BMI ≥30).

Various sensitivity analyses (as described in the “Methods” section) related to statin use remained virtually unchanged in the subgroup of patients and their controls with abdominal pain before gallstone diagnosis with cholecystectomy, in patients with or without recorded gallstone disease prior to the cholecystectomy, for the analyses in which the index dates were shifted by 30, 90, and 180 days, as well as for patients and their controls with an index date before 2004. The analysis with the total number of tablets as a proxy for days of exposure suggested that the risk of gallstone disease followed by cholecystectomy starts decreasing after about 1 to 1.5 years of treatment with statins. Additional data are available in the eFigure and eTables 1 through 7.

This large observational study provides evidence that patients with long-term statin use have a reduced risk of gallstone disease followed by cholecystectomy compared with patients without statin use. However, the OR was not decreased for patients with short-term statin use but started to decrease after 5 prescriptions, reflecting approximately 1 to 1.5 years of treatment. The risk estimate was consistent across age and sex groups. Adjustment for important risk factors^1,18,30 for gallstone disease did not materially alter the results. Diabetes mellitus was not associated with a material risk alteration³¹ so it was not included in the final multivariate model. Ischemic heart disease, ischemic stroke, and transient ischemic attack are important indications for statin use, but the findings were not altered after adjustment for these comorbidities.

The OR for gallstone disease followed by cholecystectomy for patients with long-term statin use was approximately 0.6 regardless of the comparison group chosen (ie, [presumably] patients with normal cholesterol levels or [presumably] patients with hypercholesterolemia and short-term statin use), suggesting that a possible effect of statins on gallstone formation may be largely independent of the presence of hypercholesterolemia. In addition, correction of hypercholesterolemia in patients with statin use is achieved within weeks after initiating statin therapy; in comparison, the observed risk reduction of gallstone disease followed by cholecystectomy was only evident among patients with long-term statin use (>1 to 1.5 years), suggesting contributing factors other than hypercholesterolemia. In previous studies, total cholesterol and LDL cholesterol levels have been shown to be only weakly associated with gallstone formation, if at all.^1,32- 33

The observed risk reduction of gallstone disease in patients with long-term statin use suggests a class effect for all statins, provided that the observed association is indeed causal. The stratification of current long-term use of individual statins by dose further suggested a tendency toward a lower OR for high-dose compared with low-dose exposure.

A substantially increased gallstone risk with cholecystectomy was found for patients with high BMIs and for patients with estrogen use, as well as a slightly increased risk for patients with current fibrate use, which is in accordance with the literature.^17,21 These findings may be somewhat confounded by indication because most patients take fibrates to treat high levels of triglycerides and/or low levels of high-density lipoprotein cholesterol, which are both associated with gallstone formation.^1,34- 35

Our study has several strengths. First, it is based on a large validated database with documented high data quality and completeness. Second, it encompassed more than 27 000 patients with gallstone disease followed by cholecystectomy, yielding substantial statistical power even in stratified subgroup analyses. There also was long-term follow-up for all patients, which enabled the assessment of long-term statin use, and there were sufficient data for stratification by individual statins. Third, detailed information on important comorbidities, concomitant drug therapies, and BMI was available. Finally, sensitivity analyses allowed the addressing of a potential diagnostic or protopathic bias as an alternative explanation for the observed association between statin use and the risk of gallstone disease with cholecystectomy.

On the other hand, limitations included the possibility of some outcome misclassification because original medical records were not used to validate the gallstone diagnoses and/or cholecystectomy. However, in a previous GPRD-based study, González-Pérez and García-Rodriguez¹⁸ used a similar case definition and documented a high validity of gallstone diagnoses in the GPRD. They contacted general practitioners of a random sample of 263 patients and found a 90% confirmation rate for patients who underwent cholecystectomy and an 82% confirmation rate for patients who did not have cholecystectomy. Concerning misclassification, patients in this study also were required to have had a cholecystectomy, a well-defined clinical end point mostly due to gallstone disease. In addition, outcome or exposure misclassification in observational studies tends to drive the relative risk estimates toward the null, unless misclassification is differentially distributed (eg, preferentially affecting 1 subgroup of patients), which is unlikely in this context.

Another limitation is that the analyses were not adjusted for potentially relevant lifestyle parameters altering the risk of gallstone disease such as physical activity,^36- 37 high-fat or high-carbohydrate diet,¹ intake of fruits and vegetables,³⁸ or coffee consumption.³⁹ Use of thiazide diuretics was associated with a moderately increased risk of gallbladder disease³⁰; however, inclusion of thiazide diuretics in this analysis did not alter the main association of interest. Physical activity was primarily shown to be protective in patients with normal weight or in those moderately overweight,^36- 37 while there was a strong risk reduction associated with statin use in obese patients with BMIs of 30 or higher (AOR, 0.65; 95% CI, 0.50-0.86) who are the least likely to engage in intense physical activity. A diet high in fat and carbohydrates with low intake of fruits and vegetables and low physical activity^36- 37 is markedly associated with obesity¹ and an increased risk of having gallstones. Because statin use is more frequent in overweight and obese patients (75% of all patients with statin use in the study population had a BMI ≥25), a higher gallstone risk would be expected to be associated with statin use. However, the opposite was true, suggesting that statins also are protective in overweight and obese patients.

Furthermore, there was no control for socioeconomic status. In theory, patients with low socioeconomic status may have a higher gallstone risk due to obesity, poor diet habits, and lack of physical activity. Although socioeconomic status was not found to be linked to statin prescribing in England,⁴⁰ it may be possible that patients with poor socioeconomic status are less likely to take statins because of lack of understanding or information. Therefore, poor socioeconomic status could introduce a bias toward a decreased risk of gallstone disease among patients with statin use. However, such bias should lead to a reduced gallstone risk for all statin exposure groups, and not only for those with long-term statin use. In addition, cases and controls were matched on general practice that may, to some degree, control for socioeconomic status because patients from the same neighborhood tend to consult the same general practitioner. Although a large number of confounding factors were considered and other influences were attempted to be kept to a minimum, a statement about the causality between statins and gallstone disease with cholecystectomy cannot be made due to the observational nature of the data.

In summary, this large observational study provides evidence that long-term use of statins is associated with a decreased risk of developing a diagnosis of gallstone disease requiring cholecystectomy. Our findings may be of clinical relevance given that gallstone disease represents a major burden for health care systems.