am. j. epidemiol. 2011 erichsen 162 70

American Journal of Epidemiology

ª The Author 2010. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of

Public Health. All rights reserved. For permissions, please e-mail: [email protected].

Vol. 173, No. 2

DOI: 10.1093/aje/kwq361

Advance Access publication:

November 17, 2010

Original Contribution

Long-term Statin Use and the Risk of Gallstone Disease: A Population-basedCase-Control Study

Rune Erichsen*, Trine Frøslev, Timothy L. Lash, Lars Pedersen, and Henrik Toft Sørensen

* Correspondence to Dr. Rune Erichsen, Department of Clinical Epidemiology, Aarhus University Hospital, Olof Palmes Alle 43-45,

DK-8200 Aarhus N, Denmark (e-mail: [email protected]).

Initially submitted June 17, 2010; accepted for publication September 24, 2010.

Most gallstones originate from cholesterol-supersaturated bile. Statins inhibit hepatic cholesterol biosynthesisand therefore may reduce the risk of gallstone disease. Population-based evidence, however, is sparse. Thus, theauthors conducted a population-based case-control study using medical databases from northern Denmark (1.7million inhabitants) to identify 32,494 cases of gallstones occurring between 1996 and 2008 and to identify age-,sex-, and county-matched population controls for each case. Cases and their matched controls who were exposedto lipid-lowering drugs were categorized as current users if their last prescription was redeemed �90 days beforethe case’s diagnosis date; otherwise, they were categorized as former users. Conditional logistic regression wasused to estimate adjusted odds ratios and 95% confidence intervals for gallstone disease in patients treated withlipid-lowering drugs. In current users, the adjusted odds ratios associating statin use with the occurrence ofgallstone disease were 1.17 (95% confidence interval (CI): 1.06, 1.30) for those who had 1–4 prescriptions,0.89 (95% CI: 0.80, 0.97) for those who had 5–19 prescriptions, and 0.76 (95% CI: 0.69, 0.84) for those whohad �20 total prescriptions. In former users, the corresponding adjusted odds ratios were 1.24 (95% CI: 1.11,1.39), 0.97 (95% CI: 0.86, 1.10), and 0.79 (95% CI: 0.64, 0.97), respectively. The use of other lipid-lowering drugsshowed no similar association.

cholelithiasis; gallstones; hydroxymethylglutaryl-CoA reductase inhibitors; risk

Abbreviations: AOR, adjusted odds ratio; CI, confidence interval; ICD, International Classification of Diseases; NRP, The DanishNational Registry of Patients.

The prevalence of gallstone disease in the Western worldranges from 6% to 50%, increasing with age and varyingwith geographic region (1). The risk increases with certaingenetic factors, increasing age, female gender, and obesity(1, 2). Gallstone disease is a chronic and usually asymptom-atic condition, but it leads to severe complications in about2% of symptomatic patients (3). Intense pain in the upperquadrant of the abdomen and acute cholecystitis are themost common complications and may require surgical re-moval of the gallbladder (cholecystectomy). Jaundice, pan-creatitis, and cholangitis caused by stones that migrate fromthe gallbladder to the common bile duct are other severecomplications (3). In the Western world, 80%–90% of gall-stones originate from cholesterol-supersaturated bile, withthe remainder formed as pigment stones, primarily frombilirubin and calcium (4).

In the liver, 3-hydroxy-3-methylglutaryl coenzyme A in-hibitors (statins) inhibit cholesterol biosynthesis, and thusmay prevent gallstone formation, gallstone recurrence, sub-sequent complications, and cholecystectomy. Evidence ofthe association between statins and gallstone disease is con-flicting (5), although 3 recent studies—a US cohort study (6)and case-control studies from United Kingdom (7) and Israel(8)—have shown a reduced risk of gallstone disease fol-lowed by cholecystectomy in people taking statins. Thesestudies estimated the association between statins and severegallstone disease only. In addition, the US study was basedon questionnaire reports of gallstone disease and statin use,the UK study relied on information from general practi-tioners regarding in-hospital treatments (6, 7, 9), and theIsraeli study lacked details of important covariates (8).There is, therefore, a need for evidence evaluating the

162 Am J Epidemiol 2011;173:162–170

by guest on March 17, 2015

http://aje.oxfordjournals.org/D

ownloaded from

http://aje.oxfordjournals.org/

association between statin use and the occurrence of moregeneral gallstone disease in an unselected population (10,11). To address these limitations, we conducted a study us-ing Danish population-based health registries with hospital-based information to evaluate the association between statinuse (including intensity and duration) and gallstone disease.

MATERIALS AND METHODS

Source population and setting

We conducted this population-based case-control studyfrom 1996 to 2008 using medical databases from northernDenmark, which has a population of 1.7 million (about30% of the Danish population) (12). The Danish NationalHealth Service provides tax-funded medical care for allDanish residents and partially reimburses the costs of mostphysician-prescribed drugs. We used the civil registrationnumber, a personal identifier assigned to each Danish citizenat birth and to residents at immigration, to link the Danishmedical databases (13). For each hospital admission since1977, the Danish National Registry of Patients (NRP) hasrecorded the civil registration number of the patient, datesof admission and discharge (outpatient and emergencyadmissions since 1995), surgical procedures, and up to 20discharge diagnoses coded by doctors according to theInternational Classification of Diseases (ICD) (EighthRevision until the end of 1993 and Tenth Revision thereaf-ter) (14).

Cases with gallstone disease

We identified first in- and outpatient hospitalizations inpatients with gallstone disease between January 1, 1996, andDecember 31, 2008, by searching the NRP for diagnoses ofgallstone disease (DK80), cholecytitis (DK81), or a proce-dure code for gallbladder surgery (KJKA, including chole-cystectomy and drainage). We combined these codes tocreate 1 case definition referred to as gallstone disease,and the earliest recorded date for each case was used asthe date of disease. We excluded preexisting cases of gall-stone disease by searching the NRP back to 1977 (ICD-8codes for gallstone disease: 574 and 575; procedure codesfor gallbladder surgery before 1996: 47360 and 47365). Wealso excluded cases from the NRP if the patient had a historyof liver, bile duct, or pancreatic cancer (ICD-8 codes 155–157; ICD-10 codes C22–C25) or <2 years’ history in theprescription database (see below).

Population-based controls

For each case, we selected 10 population controls with norecord of gallstone disease before the diagnosis date of thecases and matched them to the cases based on age, sex, andcounty of residence. Controls were selected from the DanishCivil Registration System, which is updated daily and hasmaintained records on vital statistics, dates of death, and theresidences of all Danish citizens and residents since April 1,1968 (13). Controls were selected using risk-set sampling,and both the cases and their matched controls were assigned

an index date identical to the date on which the case wasdiagnosed with gallstone disease.

Use of statins and other lipid-lowering drugs

Residents of northern Denmark receive prescription med-ications from pharmacies equipped with electronic account-ing systems that are primarily used to secure reimbursementfrom the National Health Service (15, 16). The pharmaciestransmit the patient’s civil registration number, the type andamount of drug prescribed according to the AnatomicalTherapeutic Chemical classification system (AppendixTable 1), and the date on which the drug was dispensed tothe prescription database. From the prescription database,we ascertained any use of statins before the index date. Wealso identified the use of fibrates, niacin, and resin, which weclassified as other lipid-lowering drugs.

We classified patients as never users of lipid-loweringdrugs (reference), current users if their last prescriptionfor statins or other lipid-lowering drugs was redeemed�90 days before the index date, or former users if their lastprescription was >90 days before the index date. Both cur-rent and former users were also categorized by total numberof prescriptions (1–4, 5–19, and �20).

Covariates

We collected data on potentially confounding factorsfrom the NRP and the prescription database (17–20). Wesearched the NRP for any hospital records containing di-agnoses of alcoholism, cancer, cardiovascular disease,chronic obstructive pulmonary disease, diabetes, hyperlip-idemia, hypothyroidism, inflammatory bowel disease, livercirrhosis, obesity, renal failure, stroke, or transient ischemicattack that occurred before the diagnosis of gallstone diseaseor the index date. We searched patient records in the pre-scription database for prescriptions for thiazides, oral anti-diabetic drugs, or postmenopausal hormone (estrogen)replacement therapy (Appendix Table 1). We coded eachof these covariates individually as a dichotomous variable(diabetes and oral antidiabetic medications were coded asa single variable).

Analysis

We calculated the frequency and proportion of cases andcontrols within categories of demographic characteristics,drug exposures, and covariates. We calculated odds ratiosand 95% confidence intervals associating drug use with theoccurrence of gallstone disease using conditional logisticregression. Because we used risk-set sampling to select con-trols, these odds ratios provided unbiased estimates of thecorresponding incidence rate ratios (21).

We calculated odds ratios adjusted for the covariates as-sociating current and former use of statins or other lipid-lowering drugs with the occurrence of gallstone disease,stratified by the total number of prescriptions. The referencegroup comprised patients with no history of using statins orother lipid-lowering drugs. These odds ratios were stratifiedby history of diabetes, cardiovascular disease, and obesity,

Statin Use and Risk of Gallstone Disease 163

Am J Epidemiol 2011;173:162–170



ownloaded from


and we used conventional logistic regression models be-cause these stratified analyses required that we dissolvethe matched sets. Therefore, we controlled for the matchedfactors as well as the other covariates.

We performed subanalyses to further explore the ob-served associations. First, we conducted 2 separate analysesin which we changed the cutoffs for current and formerusers to �30 and >30 days and �180 and >180 days, re-spectively, from the date of the last prescription to the indexdate. Second, we estimated odds ratios after changing thecategorization of statins according to the time between firstand last prescription before the index date (<1, 1–2, and �3years). We stratified this analysis by intensity of statin use.We defined intensity as a measure of the number of pillsprescribed divided by the total duration of use. Duration ofuse was determined by counting the number of days fromthe date of the first prescription to the index date. Usingthis information, we classified statin use as low-intensity(<0.5 pills/day), medium-intensity (0.5–0.8 pills/day), orhigh-intensity (�0.9 pills/day). Third, we restricted theanalysis to cases with gallbladder surgery to reduce thepotential for inclusion of false-positives. Fourth, becausethe comparison of statin users with nonusers might be con-founded by statin indications also associated with gallstonedisease (e.g., obesity, hyperlipidemia, exercise, and diet),we compared statin users with a total of 5–19 and �20prescriptions with a reference category of users with 1–4prescriptions. Because this analysis was limited to statinusers, the indications should have been more uniform. Fifth,to explore whether the associations depended on type ofstatin, we estimated odds ratios according to 7 statin prep-aration subtypes and according to lipophilicity in currentand former users (lipophile: atorvastatin, cerivastatin, flu-vastatin, lovastatin, and simvastatin; hydrophile: pravastatinand rosuvastatin). Last, we compared continuing statin users(i.e., those with �12 months between first and last prescriptionof statins and with the last prescription <12 months before theindex date) with those who discontinued statins (no pre-scription for �12 months before their index date). Becausethis analysis was limited to ever users of statins, the indica-tions should have been more uniform and included changesto diet and exercise that sometimes accompany inception ofstatin medications. Presumably, the behaviors would be lesslikely to stop at the same time as statin medication discontin-uation, so a comparison of continuing statin users with thosewho stopped statin therapy should have been less susceptibleto confounding by behavior change.

RESULTS

We identified 32,494 patients with gallstone disease(18,201 inpatients and 14,293 outpatients) and 324,925 pop-ulation controls. The median age at diagnosis/index was54.6 years, and 68.8% were women. Comorbid diseaseswere more frequent in cases than in controls (Table 1). Atotal of 1,764 cases (5.4%) with gallstone disease and(4.8%) 15,580 controls were current statin users, yieldingan overall adjusted odds ratio of 0.93 (95% confidence in-terval (CI): 0.87, 0.98). For current statin users, the adjusted

odds ratio decreased as the overall number of prescriptionsincreased, from 1.17 (95% CI: 1.06, 1.30) in patients with1–4 prescriptions to 0.76 (95% CI: 0.69, 0.84) in patientswith �20 prescriptions (Table 2). This decrease was seen inboth men and women, although women generally hadhigher odds ratios.

A total of 761 cases (2.3%) with gallstone disease and5,889 (1.8 %) controls were former statin users, which cor-responded to an adjusted odds ratio of 1.06 (95% CI: 0.98,1.16) (Table 2). The adjusted odds ratio decreased from 1.24(95% CI: 1.11, 1.39) in former users with 1–4 prescriptionsto 0.79 (95% CI: 0.64, 0.97) in former users with �20

Table 1. Characteristics of Patients With Gallstone Disease and

Population Controls in Northern Denmark, 1996–2008

Cases WithGallstoneDisease

PopulationControls

No. % No. %

Total no. 32,494 324,925

Men 10,123 31.2 101,223 31.2

Women 22,371 68.8 223,702 68.8

Age, years

0–39 8,388 25.8 83,880 25.8

40–59 11,065 34.1 110,650 34.1

60–79 9,704 29.9 97,040 29.9

�80 3,337 10.3 33,355 10.3

Time period

1996–2000 8,133 25.0 81,330 25.0

2001–2004 12,112 37.3 121,105 37.3

2005–2008 12,249 37.7 122,490 37.7

Previous hospital diagnosis

Alcoholism 814 2.5 6,283 1.9

Cancera 2,543 7.8 18,684 5.8

Cardiovascular disease 5,387 16.6 37,737 11.6

Chronic obstructivepulmonary disease

1,487 4.6 9,807 3.0

Diabetes 1,797 5.5 13,183 4.1

Hypothyroidism 348 1.1 2,381 0.7

Hyperlipidemia 277 2.2 4,967 1.5

Inflammatory boweldisease

383 1.2 2,635 0.8

Liver cirrhosis 109 0.3 488 0.2

Obesity 1,620 5.0 6,442 2.0

Renal failure 283 0.9 1,382 0.4

Stroke 1,049 3.2 8,531 2.6

Transient ischemic attack 524 1.6 4,528 1.4

Previous treatment

Hormone replacementtherapy

3,376 10.4 25,284 7.8

Thiazides 270 0.8 2,085 0.6

a All cancers except liver, bile duct, and pancreatic cancer, which

were excluded from the study.

164 Erichsen et al.

Am J Epidemiol 2011;173:162–170



ownloaded from


prescriptions. In general, female former users had a slightlyhigher risk of gallstone disease than did male former users.

Table 3 shows the odds ratios for patients with and with-out diabetes, obesity, and cardiovascular disease. Thesestratified analyses showed that the odds ratios were partic-ularly reduced in the presence of these diseases. Amongcurrent statin users, those with a previous hospital diagnosisof diabetes had an adjusted odds ratio of 0.88 (95% CI: 0.77,1.00), those with a previous hospital diagnosis of obesityhad an adjusted odds ratio of 0.80 (95% CI: 0.64, 1.00), andthose with a previous hospital diagnosis of cardiovasculardisease had an adjusted odds ratio of 0.80 (95% CI: 0.74,0.87). Our results were not impacted by a change in the splitdate for current and former users from 90 days to 30 or 180days (results not shown).

Table 4 shows that among patients with medium-term (1–2 years) and long-term (�3 years) statin use, the association

with gallstone disease was particularly protective. Further-more, as the intensity of statin use increased, the associationwith gallstones decreased.

In cases with gallbladder surgery (n¼ 744), we found thatcurrent statin users had an overall adjusted odds ratio of0.88 (95% CI: 0.62, 1.23) (1–4 prescriptions: adjusted oddsratio (AOR) ¼ 1.22, 95% CI: 0.67, 2.22; 5–19 prescriptions:AOR ¼ 0.74, 95% CI: 0.47, 1.17; and �20 prescriptions:AOR ¼ 0.78, 95 % CI: 0.46, 1.35), whereas the adjustedodds ratio for former users was 0.72 (95% CI: 0.43, 1.22)(1–4 prescriptions: AOR ¼ 0.89, 95% CI: 0.43, 1.85; 5–19prescriptions: AOR ¼ 0.99, 95% CI: 0.47, 2.08; and �20prescriptions: AOR ¼ 0.10, 95% CI: 0.01, 0.81).

In the subanalysis of statin users in which those with 1–4prescription were the reference group, we found an adjustedodds ratio of 0.76 (95% CI: 0.67, 0.86) for current statinusers with 5–19 prescriptions and an adjusted odds ratio of

Table 2. Use of Statins and the Risk of Gallstone Disease in Northern Denmark, 1996–2008

Statin Use

Cases WithGallstoneDisease

PopulationControls Crude

ORa 95% CIAdjusted

ORb 95% CI

No. % No. %

None 29,969 92.2 303,456 93.4 1.0 Reference 1.0 Reference

Current use

Overall 1,764 5.4 15,580 4.8 1.18 1.12, 1.24 0.93 0.87, 0.98

Men 741 2.3 6,733 2.1 1.14 1.05, 1.24 0.85 0.77, 0.94

Women 1,023 3.1 8,847 2.7 1.21 1.12, 1.29 0.98 0.90, 1.05

1–4 prescriptions 464 1.4 3,443 1.1 1.40 1.26, 1.54 1.17 1.06, 1.30

Men 185 0.6 1,434 0.4 1.33 1.14, 1.56 1.06 0.90, 1.25

Women 279 0.9 2,009 0.6 1.44 1.27, 1.64 1.25 1.10, 1.42

5–19 prescriptions 813 2.5 7,368 2.3 1.15 1.06, 1.24 0.89 0.82, 0.97

Men 349 1.1 3,188 1.0 1.13 1.01, 1.27 0.84 0.74, 0.95

Women 464 1.4 4,180 1.3 1.15 1.05, 1.28 0.93 0.84, 1.03

�20 prescriptions 487 1.5 4,769 1.5 1.06 0.97, 1.17 0.76 0.69, 0.84

Men 207 0.6 2,111 0.6 1.02 0.88, 1.18 0.69 0.59, 0.81

Women 280 0.9 2,658 0.8 1.10 0.97, 1.25 0.81 0.70, 0.92

Former use

Overall 761 2.3 5,889 1.8 1.35 1.24, 1.46 1.06 0.98, 1.16

Men 324 1.0 2,512 0.8 1.34 1.19, 1.51 1.00 0.88, 1.14

Women 437 1.3 3,377 1.0 1.35 1.22, 1.50 1.10 0.98, 1.22

1–4 prescriptions 353 1.1 2,408 0.7 1.52 1.36, 1.70 1.24 1.11, 1.39

Men 145 0.4 994 0.3 1.51 1.27, 1.80 1.19 0.99, 1.42

Women 208 0.6 1,414 0.4 1.53 1.32, 1.77 1.27 1.09, 1.48

5–19 prescriptions 299 0.9 2,441 0.7 1.27 1.13, 1.44 0.97 0.86, 1.10

Men 131 0.4 1,052 0.3 1.29 1.07, 1.55 0.92 0.76, 1.12

Women 168 0.5 1,389 0.4 1.26 1.07, 1.48 1.00 0.84, 1.18

�20 prescriptions 109 0.3 1,040 0.3 1.09 0.90, 1.34 0.79 0.64, 0.97

Men 48 0.1 466 0.1 1.07 0.80, 1.45 0.73 0.54, 1.00

Women 61 0.2 574 0.2 1.11 0.85, 1.45 0.83 0.63, 1.09

Abbreviations: CI, confidence interval; OR, odds ratio.a Matched based on age, sex, and county.b Controlled for the previous hospital discharges, diagnoses, and treatments mentioned in Table 1 and matched

based on age, county, sex, and period of diagnosis.


Am J Epidemiol 2011;173:162–170



ownloaded from


0.65 (95% CI: 0.56, 0.74) for current statin users with �20prescriptions. For former statin users, the correspondingadjusted odds ratios were 0.78 (95% CI: 0.66, 0.92) and0.64 (95% CI: 0.51, 0.80), respectively. In the final subanal-ysis, we found that statin users with no statin prescription �12months before their index date had an adjusted odds ratioof 1.39 (95% CI: 1.19, 1.61) compared with continuingstatin users.

When we divided the cases into the 7 statin preparationsubtypes and also according to lipophilicity, we found nomajor heterogeneity of the odds ratios (results not shown).In the analysis of other lipid-lowering drugs, we identified70 current users (0.2%) and 122 former users (0.4%) amongcases and 377 current users (0.1%) and 775 former users(0.2%) among controls. For current users, this correspondedto an adjusted odds ratio of 1.42 (95% CI: 1.09, 1.84) (stratifiedby the number of prescriptions: 1–4: AOR ¼ 2.16, 95% CI:1.06, 4.40; 5–19: AOR ¼ 1.11, 95% CI: 0.67, 1.84; and �20:AOR ¼ 1.46, 95% CI: 1.05, 2.05); for former users, it corre-sponded to an adjusted odds ratio of 1.18 (95% CI: 0.97, 1.43)(stratified on the number of prescriptions: 0–4: AOR ¼1.20, 95% CI: 0.89, 1.62; 5–19: AOR ¼ 0.96, 95% CI:0.67, 1.38; and �20: AOR ¼ 1.44, 95% CI: 1.01, 2.07).

DISCUSSION

In this population-based case-control study, current statinusers had a decreased odds ratio for gallstone disease com-pared with nonusers, so long as they had �5 prescriptions.This duration corresponds to 1–2 years of statin use, depen-dent upon the intensity of the use. For former users, thedecrease in odds ratios was less pronounced and presentonly after �20 prescriptions, but the sustained decrease inodds ratios in these patients may reflect a long-standingeffect of statins in the prevention of gallstone disease. Thispattern of results was insensitive to type or lipophilicity ofstatins, changes in the definition of current and former users(30 or 180 days vs. 90 days), and the restriction to cases withgallbladder surgery. In addition, patients who discontinuedstatin treatment �12 months before their index date had anincreased odds ratio for gallstone disease compared withcontinuing statin users. As expected, we found no reductionin the odds ratio for gallstone disease in patients treated withother lipid-lowering drugs.

Our results extend the results of other studies (6–8, 22–24). In 2009, Bodmer et al. (7) published a study that in-cluded 27,035 cases with gallstone disease treated with

Table 3. Use of Statins and the Risk of Gallstone Disease in the Presence or Absence of

Diabetes, Obesity, and Cardiovascular Disease in Northern Denmark, 1996–2008

CrudeORa 95% CI

AdjustedORb 95% CI

Previous hospital diagnosisof diabetes

Current statin users 0.96 0.85, 1.09 0.88 0.77, 1.00

Former statin users 1.15 0.98, 1.37 1.05 0.88, 1.25

No previous hospital diagnosisof diabetes



Previous hospital diagnosisof obesity



No previous hospital diagnosisof obesity



Previous hospital diagnosisof cardiovascular disease



No previous hospital diagnosisof cardiovascular disease



Abbreviations: CI, confidence interval; OR, odds ratio.a Controlled for age, county, sex, and period of diagnosis.b Controlled for the previous hospital discharges, diagnoses, and treatments mentioned in

Table 1 and matched on age, county, sex, and period of diagnosis. The reference group was

the number of statin users in each stratum.

166 Erichsen et al.

Am J Epidemiol 2011;173:162–170



ownloaded from


cholecystectomy and 106,531 controls. The study was basedon the UK General Practice Research Database, which wasrepresentative of the United Kingdom population in mostaspects (9), and was supported by results from another case-control study from Israel (8). Bodmer et al. found adjustedodds ratios similar ours (e.g., current users with �20 pre-scriptions had an adjusted odds ratio of 0.67 (95% CI: 0.55,0.80)), and they had to some extent better confounder con-trol (e.g., they controlled for body mass index and smokingstatus, although those data were ascertained close to theindex date, rather than before the first statin prescription).However, the study (7) only included patients with severegallstone-related complications and relied on informationfrom general practitioners regarding the in-hospital proce-dure of cholecystectomy.

Tsai et al. (6) published a cohort study in 2009 that in-cluded 2,479 women from the Nurses’ Health Study whohad undergone cholecystectomy. The study included infor-mation on statin use, cholecystectomy, and other covariatesfrom questionnaires and found a decreased risk of cholecys-tectomy after statin use (multivariate relative risk ¼ 0.82, 95%CI: 0.70, 0.96). In contrast, a 2001 French cross-sectionalstudy by Caroli-Bosc et al. (5) that included 830 patientsundergoing gallbladder ultrasonography (of whom only4.6% reported a previous statin use) did not find any asso-ciation between statin use and gallstone disease. Moreover,

reports on alterations in human bile cholesterol saturationafter statin therapy showed compelling evidence that indi-cated that statins decreased the cholesterol saturation of bile(23, 24), although results of a few reports did not confirmthis association (25, 26). Animal studies have also reportedbeneficial effects of statins in the prevention of gallstoneformation (27, 28).

The strengths of our study include the population-baseddesign within a universal health care system with completehospital and prescription history and access to appropriatepopulation controls. In addition, our analysis adjusted forseveral important risk factors for gallstone disease.

Our study also has limitations. We used redeemed pre-scriptions as a surrogate for actual statin consumption,which may be imperfect. However, we stratified our analy-ses on the overall number of prescriptions, and therefore thepossibility of including nonusers, at least among patientswith several prescriptions, is minimal. The categorizationof current and former statin users is also imperfect, and thusthere might be current users in the group of former users andvice versa. The subanalysis in which we changed the splitdate for current and former users, however, supported theresults of our main analysis. Furthermore, registry diagnosesmay be inaccurate, and thus misclassification of gallstonedisease could occur, although the overall quality and accu-racy of the data in the NRP has been estimated to be high

Table 4. Lengtha and Intensityb of Statin Use and the Risk of Gallstone Disease in Northern

Denmark, 1996–2008

Statin UseCrudeORc 95% CI

AdjustedORd 95% CI

None 1.0 Reference 1.0 Reference

Short-term (<1 year)

Overall 1.46 1.36, 1.57 1.21 1.12, 1.30

Low-intensity (<0.5 pills/day) 1.55 1.35, 1.79 1.26 1.09, 1.45

Medium-intensity (0.5–0.8 pills/day) 1.41 1.19, 1.66 1.18 1.00, 1.39

High-intensity (�0.9 pills/day) 1.44 1.31, 1.58 1.20 1.09. 1.32

Medium-term (1–2 years)

Overall 1.14 1.05, 1.23 0.89 0.82, 0.97



High-intensity (�0.9 pills/day) 1.09 0.98, 1.21 0.84 0.76, 0.94

Long-term (�3 years)

Overall 1.10 1.02, 1.18 0.80 0.73, 0.86



High-intensity (�0.9 pills/day) 1.07 0.98, 1.18 0.77 0.70, 0.85

Abbreviations: CI, confidence interval; OR, odds ratio.a Short-, medium-, and long-term use were calculated using the time from first prescription to

last prescription before date of gallstone disease or index date.b Intensity was determined as number of pills prescribed divided by the time from first pre-

scription to the date of gallstone disease or index date separated into 3 categories.c Controlled for age, county, sex, and period of diagnosis.d Controlled for the previous hospital discharges, diagnoses, and treatments mentioned in

Table 1 and matched on age, county, sex, and period of diagnosis.


Am J Epidemiol 2011;173:162–170



ownloaded from


(29). The subanalysis of cases with gallbladder surgery wasconducted to increase the positive predictive value of thediagnosis of gallstone disease (30), and it confirmed ourmain finding. Moreover, we find it unlikely that misclassi-fication of gallstone disease would be unequally distributedbetween statin users and nonusers, and thus any such mis-classification would bias estimates of dichotomous exposureassociations towards the null. However, the problem of mis-classification would become evident in the stratified analysisof patients with and without a diagnosis of obesity (i.e., anobesity diagnosis registered in the NRP has a high positivepredictive value but a low sensitivity). Hence, the odds ra-tios for gallstone disease in obese patients would reflect theratios of the very obese, whereas those of patients withoutobesity would be measured by the overall estimate.

Diagnoses of asymptomatic or mild gallstone diseasemay be more likely among statin users than among nonusersbecause of the routine physical examination of statin users.This tendency would cause overestimation of the risk ofgallstone disease among statin users compared with non-users, and therefore statins may not explain the decreasedodds ratios. Furthermore, the subanalysis of patients withgallbladder surgery (i.e., symptomatic cases), in which itseems unlikely that statin users would have a higher chanceof being diagnosed, confirmed our main finding.

Another limitation is that the exposure group of statinusers presumably differs from nonusers in aspects con-nected to an increased risk of gallstone disease, such as poordiet and physical inactivity. Hence, our results could havebeen confounded by indication, because new users of statinsmight also have improved their health behaviors. However,the results for other lipid-lowering drugs showed no de-creasing risk of gallstone disease, although these associa-tions would have been susceptible to the same confounding.Moreover, although we adjusted for differences in severalimportant gallstone risk factors (or surrogates thereof), theremay have been unmeasured confounding. This unmeasuredconfounding would arise from factors on which we did nothave certain information, such as physical activity levels,socioeconomic status (although not strongly associated withstatin use in Denmark (31)), and levels of triglycerides, aswell as from poorly measured factors, such as obesity. How-ever, these confounding factors all tend to be associated withan increased risk of gallstone disease, and thus may explainthe increased risk of gallstone disease for statin users with1–4 prescriptions but not the risk-reducing association withstatin use of longer duration. Furthermore, by comparingstatin users with 5–19 and �20 prescriptions with statinusers with 1–4 prescriptions, we were able to reduce thepotential for bias from some of these confounders, as theywould be more similarly distributed in these 3 statin expo-sure groups than in a comparison between statins users andnever users. A similar situation is present in the stratifiedanalysis based on history of cardiovascular disease. In pa-tients with a history of cardiovascular disease, these con-founding factors were probably relatively equallydistributed between cases and controls and thereby limitedimpact on the odds ratio, whereas they were more unequallydistributed in patients with no history of cardiovasculardisease.

As noted above, the initiation of statin treatment is some-times also accompanied by changes in health behaviors,such as increased physical activity and eating a healthierdiet. It is therefore possible that the decreased risk of gall-stone disease associated with initiation of statin use wascaused not only by statins but also by the beginning ofhealth-seeking behaviors. Although we cannot rule out thisconfounding, our analysis comparing cases who discontin-ued statins with continuing statin users, which showeda higher risk among those who stopped statin therapy, sup-ported the hypothesis that statins in fact do reduce the risk ofgallstone disease directly. Furthermore, lifestyle factorswere not found to be important confounders in the studyby Tsai et al. (6) or the study by Bodmer et al. (7, 32).

It is well-known that fibrates (included in the group ofother lipid-lowering drugs in this study) increase the risk ofgallstone disease (5), and our study supported this associa-tion. Because the residual and unmeasured confounding dis-cussed above is expected to cause overestimation of theassociation between gallstone disease and these drugs, ourresults should not be taken alone as evidence of a potentiallyharmful effect. In conclusion, this population-based case-control study confirmed that long-term sustained statin usedecreases the risk of gallstone disease in both men and women.

ACKNOWLEDGMENTS

Author affiliations: Department of Clinical Epidemiol-ogy, Aarhus University Hospital, Aarhus, Denmark (RuneErichsen, Trine Frøslev, Timothy L. Lash, Lars Pedersen,Henrik Toft Sørensen); and Department of Epidemiology,Boston University School of Public Health, Boston, Massa-chusetts (Timothy L. Lash, Henrik Toft Sørensen).

This work was supported by the Clinical EpidemiologyResearch Foundation, Aarhus University Hospital, Aarhus,Denmark.

The preliminary findings of this study were presented asa poster at the 26th International Conference on Pharmaco-epidemiology and Theraputic Risk Management, Brighton,United Kingdom, August 19–22, 2010, and published inabstract form (33).

The Department of Clinical Epidemiology at Aarhus Uni-versity Hospital receives funding for other studies fromcompanies in the form of research grants to (and adminis-tered by) Aarhus University. None of these studies have anyrelation to the present study.

Conflict of interest: none declared.

REFERENCES

1. Kratzer W, Mason RA, Kachele V. Prevalence of gallstones insonographic surveys worldwide. J Clin Ultrasound. 1999;27(1):1–7.

2. Katsika D, Grjibovski A, Einarsson C, et al. Genetic and en-vironmental influences on symptomatic gallstone disease:a Swedish study of 43,141 twin pairs. Hepatology. 2005;41(5):1138–1143.

168 Erichsen et al.

Am J Epidemiol 2011;173:162–170



ownloaded from


3. Portincasa P, Moschetta A, Petruzzelli M, et al. Gallstonedisease: symptoms and diagnosis of gallbladder stones. BestPract Res Clin Gastroenterol. 2006;20(6):1017–1029.

4. Portincasa P, Moschetta A, Palasciano G. Cholesterol gallstonedisease. Lancet. 2006;368(9531):230–239.

5. Caroli-Bosc FX, Le Gall P, Pugliese P, et al. Role of fibratesand HMG-CoA reductase inhibitors in gallstone formation:epidemiological study in an unselected population. Dig DisSci. 2001;46(3):540–544.

6. Tsai CJ, Leitzmann MF, Willett WC, et al. Statin use and therisk of cholecystectomy in women. Gastroenterology. 2009;136(5):1593–1600.

7. Bodmer M, Brauchli YB, Krahenbuhl S, et al. Statin use andrisk of gallstone disease followed by cholecystectomy. JAMA.2009;302(18):2001–2007.

8. Merzon E, Weiss NS, Lustman AJ, et al. Statin administrationand risk of cholecystectomy: a population-based case-controlstudy. Expert Opin Drug Saf. 2010;9(4):539–543.

9. Walley T, Mantgani A. The UK General Practice ResearchDatabase. Lancet. 1997;350(9084):1097–1099.

10. Fujita T. Use of statins and gallstone risk. JAMA. 2010;303(12):1146–1147.

11. Hsu CS, Kao JH. Use of statins and gallstone risk [letter].JAMA. 2010;303(12):1146.

12. StatBank Denmark. Population and Elections. Table FOLK1.Copenhagen, Denmark: Statistics Denmark; 2010. (http://www.statistikbanken.dk/statbank5a/default.asp?w¼1613).(Accessed on November 1, 2010).

13. Frank L. Epidemiology. When an entire country is a cohort.Science. 2000;287(5462):2398–2399.

14. Andersen TF, Madsen M, Jørgensen J, et al. The Danish Na-tional Hospital Register. A valuable source of data for modernhealth sciences [in Danish].DanMed Bull. 1999;46(3):263–268.

15. Nielsen G, Sørensen H, Zhou W, et al. The pharmacoepide-miologic prescription database of North Jutland—a valid toolin pharmacoepidemiological research. Int J Risk Safety Med.1997;10(3):203–205.

16. Gaist D, Sørensen HT, Hallas J. The Danish prescription reg-istries [in Danish]. Dan Med Bull. 1997;44(4):445–448.

17. Kratzer W, Haenle MM, Mason RA, et al. Prevalence ofcholelithiasis in patients with chronic inflammatory boweldisease. World J Gastroenterol. 2005;11(39):6170–6175.

18. Attili AF, Capocaccia R, Carulli N, et al. Factors associatedwith gallstone disease in the MICOL experience. MulticenterItalian Study on Epidemiology of Cholelithiasis. Hepatology.1997;26(4):809–818.

19. Mendez-Sanchez N, Bahena-Aponte J, Chavez-Tapia NC,et al. Strong association between gallstones and cardiovasculardisease. Am J Gastroenterol. 2005;100(4):827–830.

20. Yoo EH, Lee SY. The prevalence and risk factors for gallstonedisease. Clin Chem Lab Med. 2009;47(7):795–807.

21. Rothman KJ, Greenland S, Lash TL, eds. Modern Epidemiol-ogy. 3rd ed. Philadelphia, PA: Lippincott Williams & Wilkins;2008.

22. Smith JL, Roach PD, Wittenberg LN, et al. Effects of sim-vastatin on hepatic cholesterol metabolism, bile lithogenicityand bile acid hydrophobicity in patients with gallstones.J Gastroenterol Hepatol. 2000;15(8):871–879.

23. Reihner E, Rudling M, Stahlberg D, et al. Influence ofpravastatin, a specific inhibitor of HMG-CoA reductase, onhepatic metabolism of cholesterol. N Engl J Med. 1990;323(4):224–228.

24. Duane WC, Hunninghake DB, Freeman ML, et al. Simvasta-tin, a competitive inhibitor of HMG-CoA reductase, lowerscholesterol saturation index of gallbladder bile. Hepatology.1988;8(5):1147–1150.

25. Smit JW, van Erpecum KJ, Renooij W, et al. The effects of the3-hydroxy, 3-methylglutaryl coenzyme A reductase inhibitorpravastatin on bile composition and nucleation of cholesterolcrystals in cholesterol gallstone disease. Hepatology. 1995;21(6):1523–1529.

26. Tazuma S, Ohya T, Mizuno T, et al. Effects of fluvastatinon human biliary lipids. Am J Cardiol. 1995;76(2):110A–113A.

27. Davis KG, Wertin TM, Schriver JP. The use of simvastatin forthe prevention of gallstones in the lithogenic prairie dogmodel. Obes Surg. 2003;13(6):865–868.

28. Abedin MZ, Narins SC, Park EH, et al. Lovastatin altersbiliary lipid composition and dissolves gallstones: a long-term study in prairie dogs. Dig Dis Sci. 2002;47(10):2192–2210.

29. Mosbech J, Jørgensen J, Madsen M, et al. The national patientregistry. Evaluation of data quality [in Danish]. Ugeskr Laeger.1995;157(26):3741–3745.

30. Thomsen RW, Thomsen HF, Nørgaard M, et al. Risk ofcholecystitis in patients with cancer: a population-basedcohort study in Denmark. Cancer. 2008;113(12):3410–3419.

31. Thomsen RW, Johnsen SP, Olesen AV, et al. Socioeconomicgradient in use of statins among Danish patients: population-based cross-sectional study. Br J Clin Pharmacol. 2005;60(5):534–542.

32. Bodmer M, Brauchli YB, Meier CR. Use of statins and gall-stone risk—reply [letter]. JAMA. 2010;303(12):1147.

33. Erichsen R, Frøslev T, Pedersen L, et al. Long-term statin usereduces the risk of gallstone disease: a population-based case-control study [abstract]. Pharmacoepidemiol Drug Safe.2010;19(suppl 1):138.

(Appendix Table 1 follows)


Am J Epidemiol 2011;173:162–170



ownloaded from

http://www.statistikbanken.dk/statbank5a/default.asp?w=1613




Appendix Table 1. International Classification of Diseases Codes

and Anatomical Therapeutic Chemical Classification System Codes

used in the Analysis

Hospital discharge codes

Alcoholism

ICD-10: F10 (except F10.0), G31.2, G62.1, G72.1, I42.6,K29.2, K86.0, Z72.1

ICD-8: 291, 303, 577.10, 571.09, 571.10

and/or a prescription of disulfiram (ATC N07BB01)

Apoplexia cerebri

ICD-10: I61, I63, I64

ICD-8: 431, 433, 434

Chronic obstructive pulmonary disease

ICD-10: J41, J42, J43, J44

ICD-8: 491, 492

Cancer

ICD-10: C0–C97 (except C22–25)

ICD-8: 14–20 (except 155, 156, 157)

Cardiac disease

ICD-10: I05–I09, I10–I15, I20–I25, I50

ICD-8: 393–398, 400–404, 410–414, 427.09, 427.10, 427.19

Diabetes

ICD-10: E10.0, E10.1–E10.9, E11.0; E11.1–E11.9

ICD-8: 249.00, 249.01– 249.09, 250.00, 250.01–250.09

Hypothyroidism

ICD-10: E00–E03

ICD-8: 244

Hyperlipidemia

ICD-10: E78.0–E78.5

ICD-10: 272.00, 272.01, 279.00, 279.01

Inflammatory bowel disease

ICD-10: K50, K51

ICD-8: 563.01, 563.19

Table continues

Appendix Table 1. Continued

Liver cirrhosis

ICD-10: K70.3, K74.6

ICD-8: 571.09, 571.92, 571.99

Obesity

ICD-10: E66

ICD-8: 277

Renal failure

ICD-10: N17–N19

ICD-8: 581–584

Transient ischemic attack

ICD-10: G45.9

ICD-8: 435

Anatomical Therapeutic Chemical Classification System

Statins

Atorvastatin C10AA05

Cerivastatin C10AA06

Fluvastatin C10AA04

Lovastatin C10AA02

Pravastatin C10AA03

Rosuvastatin C10AA07

Simvastatin C10AA01

Other lipid-lowering drugs

Fibrates C10AB

Niacins C10AD

Resins C10AC

Other drugs (covariates)

Hormone replacement therapy: G03CA, G03FA

Thiazide: CO3AA

Oral antidiabetics: A10

Abbreviations: ICD-8, International Classification of Diseases,

Eighth Revision; ICD-10, International Classification of Diseases,

Tenth Revision.

170 Erichsen et al.

Am J Epidemiol 2011;173:162–170



ownloaded from


am. j. epidemiol. 2011 erichsen 162 70

Documents