Risk factors for intrahepatic cholangiocarcinoma in a low-risk population: A nationwide case-control study


  • Tania M. Welzel,

    Corresponding author
    1. Division of Cancer Epidemiology and Genetics, Department of Health and Human Services, National Cancer Institute, National Institutes of Health, Bethesda, MD
    • Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD 20892-7234, USA
    Search for more papers by this author
  • Lene Mellemkjaer,

    1. The Danish Cancer Society, Institute of Cancer Epidemiology, Copenhagen, Denmark
    Search for more papers by this author
  • Gridley Gloria,

    1. Division of Cancer Epidemiology and Genetics, Department of Health and Human Services, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Lori C. Sakoda,

    1. Division of Cancer Epidemiology and Genetics, Department of Health and Human Services, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Ann W. Hsing,

    1. Division of Cancer Epidemiology and Genetics, Department of Health and Human Services, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Laure El Ghormli,

    1. Division of Cancer Epidemiology and Genetics, Department of Health and Human Services, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author
  • Jorgen H. Olsen,

    1. The Danish Cancer Society, Institute of Cancer Epidemiology, Copenhagen, Denmark
    Search for more papers by this author
  • Katherine A. McGlynn

    1. Division of Cancer Epidemiology and Genetics, Department of Health and Human Services, National Cancer Institute, National Institutes of Health, Bethesda, MD
    Search for more papers by this author


Recently, the incidence of intrahepatic cholangiocarcinoma (ICC) has been increasing in a number of developed (Western) countries. However, risk factors in these low-risk populations are poorly understood. In this nationwide population based case-control study in Denmark, we examined the relationship between selected medical conditions and subsequent ICC risk to provide additional clues to etiopathogenesis. All histologically confirmed ICC cases diagnosed in Denmark between 1978 and 1991 were identified from the Danish cancer registry. Population controls were selected from the central population registry and were matched 4:1 to cases on sex and year of birth. Cases and controls were linked to the Danish hospital discharge registry to obtain information on prior hospital diagnoses. Odds ratios (OR) and 95% confidence intervals (95% CI) were derived using conditional logistic regression. A total of 764 ICC cases and 3,056 population controls were included in the study. Chronic liver diseases were significantly related to ICC: alcoholic liver disease (OR = 19.22, 95% CI = 5.55–66.54), unspecified cirrhosis (OR = 75.9, 95% CI 10.2–565.7). Bile duct diseases were also associated with risk: cholangitis (OR = 6.3, 95% CI = 2.3–17.5), choledocholithiasis (OR = 23.97, 95% CI = 2.9–198.9), cholecystolithiasis (OR = 4.0, 95% CI = 2.0–7.99), though gallbladder removal did not change risk (OR = 1.6, 95% CI = 0.65–3.7). Among other conditions, chronic inflammatory bowel disease (OR = 4.7, 95% CI = 1.65–13.9) was significantly associated with ICC. Diabetes was associated with risk in the year prior to diagnosis of ICC (OR = 3.02, 95% CI = 1.05–8.69). Obesity was unrelated to risk. These results confirm that prior bile duct diseases increase risk of ICC and suggest that alcoholic liver disease and diabetes may also increase risk. © 2006 Wiley-Liss, Inc.

Intrahepatic cholangiocarcinoma (ICC), the second most common form of primary liver cancer (after hepatocellular carcinoma) in the world, is characterized by wide variability in incidence and risk factors.1 In high-risk areas, such as Northeast Thailand, the primary risk factor for ICC is infestation with liver flukes (Ophistorchis viverrini, Clonorchis sinensis and hepatolithiasis).2, 3, 4 In low-risk areas, such as Europe and North America, ICC incidence has been increasing in recent decades5, 6 and the etiology is poorly understood as the conduct of epidemiological studies is complicated by the rarity of the tumors. In these areas, primary sclerosing cholangitis and inflammatory bowel diseases are the most commonly reported risk factors for ICC.7, 8, 9, 10 It is unclear whether the incidence of these conditions has changed, however, and, as these conditions are generally more common among younger persons with ICC,10 they are unlikely to account for increasing ICC risk in the elderly population. Additionally, alcoholic liver disease, cirrhosis, hepatitis C virus, human immunodeficiency virus and diabetes have recently been reported to be associated with the development of ICC in the United States.11 Whether these pre-existing medical conditions are also associated with ICC in other low-risk populations has not been previously reported.

To provide additional clues to ICC etiology, we examined the relationship of selected medical conditions to ICC in a large population-based record linkage study in Denmark using data from the Danish cancer registry.

Material and methods

All persons diagnosed with ICC in Denmark between 1978 and 1991 were identified in the Danish cancer registry. The Danish cancer registry was initiated in 1943 as a population-based country-wide registry.12 Newly diagnosed cancers, as well as subsequent changes of the initial diagnosis, are routinely reported to the registry through hospitals, outpatient clinics and general practitioners. Pathology reports, as well as autopsy notifications including incidental cancer findings, and death certificates are also provided to the registry. Records for each patient include name, central population identification number, date of birth, residence at date of cancer diagnosis, method of verification of the cancer, and extent of disease at diagnosis and treatment. To further increase the completeness of data, the registry has been periodically linked to the national hospital discharge registry since 1987, and missing data are supplemented. Since January 1, 1978 cancers are classified according to the Danish version of the International Classification of Diseases, 7th revision13 and additionally according to the International Classification of Diseases for Oncology (ICD-O).14 Completeness of the Danish cancer registry is 95–98% and diagnostic validity is high.12

To avoid inclusion of other types of liver cancer, such as hepatocellular carcinoma, or metastatic liver cancer, only patients with histological or cytological confirmation of ICC (ICD-O topography C22.0 (liver), C22.1 (intrahepatic bile duct) and histology code 8160) were selected for the study. Patients with diagnoses of prior cancers were excluded, as were Danish citizens residing in Greenland and the Faroe Islands.

Population controls were identified from the Danish central population register which was established in 1968. This registry assigns all Danish citizens a unique personal identification number that encodes gender and date of birth, and keeps information on residence, emigration and vital status. The identification number is widely applied in the Danish health care system and permits the linkage of each citizen's information collected in different national registries. The controls were randomly chosen among inhabitants alive and free of cancer (except non-melanoma skin cancer) on the year of cancer diagnosis for the matched case. The controls were matched 4:1 to the cases on sex and year of birth. To determine prior hospital diagnoses, cases and controls were linked to the Danish hospital discharge registry via their personal identification number. The Danish hospital discharge registry (also known as, the Danish national registry of patients) is a centralized registry that has collected information on virtually all hospital discharges in Denmark since 1977 (The Danish national board of health, 1981).15 Each admission record includes the person's personal identification number, dates of admission and discharge, and up to 20 discharge diagnoses and information on surgical procedures. Diagnoses are classified according to the Danish version of the International Classification of Diseases, 8th Version (ICD-8)16 and surgical procedures according to a Danish classification of surgical procedures and therapies (operation-codes, Danish national board of health classification of surgical procedures and therapies).17

The relationships of ICC to possible risk factors in 3 broad categories were examined. The first category, chronic non-infectious liver diseases, was composed of alcoholic liver disease (alcoholic steatosis and alcoholic cirrhosis; ICD-codes 571.09 and 571.10), and nonspecific cirrhosis (571.92) which was defined as cirrhosis in the absence of any alcohol related codes (ICD code 303). The second category, biliary tract diseases, included cholangitis (575.04), choledocholithiasis (574.03, 574.04), cholecystolithiasis (574.00, 574.09) and the surgical procedure cholecystectomy (op-code 4736). Because of the absence of a specific code in ICD-8, primary sclerosing cholangitis could not be investigated independently of all cholangitis. The third category, other risk factors, consisted of diabetes mellitus (249, 250), obesity (277.99) and chronic inflammatory bowel diseases (563).

The 3 categories of medical conditions were compared among ICC patients and population controls using conditional logistic regression to calculate odds ratios and 95% confidence intervals. All analyses were adjusted for the matching factors (i.e., age at ICC diagnosis, sex and year of birth). Tests of statistical significance and confidence intervals were two-sided, and p < 0.05 was considered statistically significant. The analyses were conducted using SAS, Version 9.1 (SAS Institute, Cary NC, USA).


During the study period, 849 patients with histologically confirmed ICC were reported to the Danish cancer registry and 764 (90%) of these met our inclusion criteria. The control group consisted of 3,056 age and sex matched individuals without ICC or any other cancer diagnosis. The characteristics of the study participants are summarized in Table I. As the cases and controls were individually matched on year of birth and sex, there were no differences in the distributions of these variables. Approximately 41% were aged 75 years or older at the time of diagnosis, and 54% of the study participants were female. The majority of the ICC cases were diagnosed in the 1980s (74%), while ∼15% were diagnosed in the 1970s and 11% in the 1990s.

Table I. Characteristics of the ICC Cases and Population Controls
 ICC cases (n = 764)Controls (n = 3056)
  • 1

    For the controls, year of diagnosis equals reference year.

Age at diagnosis (years)
 <65218 (28.5%)872 (28.5%)
 65–74229 (30.0%)916 (30.0%)
 75+317 (41.5%)1,268 (41.5%)
 Male355 (46.5%)1,420 (46.5%)
 Female409 (53.5%)1,636 (53.5%)
Year of diagnosis1
1978–1979117 (15.3%)468 (15.3%)
1980–1984276 (36.1%)1,104 (36.1%)
1985–1989290 (38.0%)1,160 (38.0%)
1990–199181 (10.6%)324 (10.6%)

Table II shows the odds ratios (ORs) and 95% confidence intervals (95% CIs) associated with the medical conditions of interest according to the 3 broad groupings. Also shown are the ORs and 95% CIs of each condition stratified by the time between the diagnosis of the condition and the ICC diagnosis (i.e. ≤1 year before diagnosis of ICC vs. >1 year before the diagnosis of ICC).

Table II. Odds Ratios and 95% Confidence Intervals of Selected Diagnoses or Procedures Associated with Subsequent ICC Risk
 ICC cases (n = 764)Controls (n = 3056)OR95% CI
Chronic liver diseases
Alcoholic liver diseases
 No749(98.04)305299.971.00 (ref)
  ≤1 year7(0.92)0
  >1 year8(1.05)4(0.13)10.672.83–40.21
Nonspecific cirrhosis
 No745(97.51)3055(99.97)1.00 (ref)
  ≤1 year15(1.96)1(0.03)60.007.93–454.22
  >1 year4(0.52)0
Bile duct diseases
 No754(98.69)3049(99.77)1.00 (ref) 
  ≤1 year4(0.52)2(0.07)8.001.47–43.68
  >1 year6(0.79)5(0.16)5.511.54–19.72
 No758(99.21)3055(99.97)1.00 (ref) 
  ≤1 year3(0.39)0
  >1 year3(0.39)1(0.03)12.001.25–115.36
 No747(97.77)3038(99.41)1.00 (ref) 
  ≤1 year8(1.05)3(0.10)11.012.92–41.59
  >1 year9(1.18)15(0.49)2.571.10–5.99
 No757(99.08)3038(99.41)1.00 (ref) 
  ≤1 year1(0.13)3(0.10)1.330.14–12.82
  >1 year6(0.79)15(0.49)1.600.62–4.13
Other Conditions
Inflammatory bowel disease
 No757(99.08)3050(99.80)1.00 (ref) 
  ≤1 year2(0.26)1(0.03)8.000.73–88.23
  >1 year5(0.65)5(0.16)4.001.16–13.82
 No749(98.03)3013(98.59)1.00 (ref) 
  ≤1 year6(0.79)8(0.26)3.021.05–8.69
  >1 year9(1.18)35(1.15)1.060.50–2.24
 No758(99.21)304499.611.00 (ref) 
  ≤1 year2(0.26)3(0.10)2.670.45–15.96
  >1 year4(0.52)9(0.29)1.830.54–6.14

Chronic liver diseases were strongly associated with ICC. Alcoholic liver disease (alcoholic steatosis and alcoholic cirrhosis), as well as unspecified liver cirrhosis, were more frequent in cases (OR = 19.22, 95% CI = 5.55–66.54 and OR = 75.86, 95% CI = 10.17–565.68, respectively) compared to controls in the overall analysis, as well as in the analyses excluding diagnoses in the year prior to ICC diagnosis (Table II).

Among bile duct diseases, cases were more likely to have had a prior diagnoses of cholangitis (OR = 6.32, 95% CI = 2.29–17.48), choledocholithiasis (OR = 23.97, 95% CI = 2.89–198.97) and cholecystolithiasis (OR = 4.02, 95% CI = 2.02–7.99). All bile duct conditions remained significantly more frequent in the overall analysis, than when the analysis was restricted to conditions first diagnosed more than 1 year prior to ICC diagnosis (Table II).

Previous cholecystectomy was more common in ICC cases compared to controls in the overall analysis (OR = 1.56, 95% CI = 0.65–3.73) as well as in the stratified analysis, (OR = 1.33, 95% CI = 0.14–12.82; OR = 1.60, 95% CI = 0.62–4.13), however, the results in neither instance attained statistical significance. In addition, inflammatory bowel disease was significantly associated with ICC (OR = 4.67, 95% CI = 1.57–13.89). In the stratified analysis, inflammatory bowel disease was more frequent in cases than controls during both time intervals, though the relationship only attained statistical significance in the interval >1 year (OR = 4.00, 95% CI = 1.16–13.82).

Although the prevalence of diabetes mellitus (OR = 1.43, 95% CI = 0.78–2.63) and obesity (OR = 2.05, 95% CI = 0.75–5.58) was higher in ICC cases compared to controls, the relationship did not attain statistical significance in the overall analysis. Similarly, diabetes mellitus and obesity did not differ between cases and controls in the interval of >1 year prior to ICC diagnosis (OR = 1.06, 95% CI = 0.5–2.24 and OR = 1.83, 95% CI = 0.54–6.14), however, diabetes was 3 times more frequently diagnosed in cases within 1 year prior to ICC diagnosis (OR = 3.02, 95% CI = 1.05–8.69).


This is the largest nationwide European population-based case-control study to examine risk factors for ICC. The present study comprises all patients with histologically confirmed ICC in Denmark over a time period of 14 years and confirms the association between ICC and such well-established risk factors as cholangitis and inflammatory bowel disease, which frequently is associated with cholangitis.7 Chronic inflammation of the bile ducts as well as cholestasis may contribute to the malignant transformation of the biliary epithelium in patients with PSC, and the prevalence of ICC in patients with PSC ranges from ∼7–13%.18

In addition, chronic liver diseases such as alcoholic liver disease and unspecified cirrhosis were significantly more common in ICC cases compared to controls, in parallel to what has been established for hepatocellular carcinoma (HCC). These findings are in agreement with a previous US case-control study examining ICC risk factors in an elderly population. In that study of 625 cases and 90,843 population controls, a greatly increased risk associated with alcoholic liver disease or unspecified cirrhosis was reported.11 Alcohol may be related to ICC in a similar manner to the way it is related to HCC, as there is increasing evidence that both types of primary liver cancers may share common progenitor cells.19 Alcohol-induced oxidative stress may create an environment that is conducive to alcohol-mediated liver injury and the development of both HCC and ICC.20, 21 In the present study, some of the ICC cases had pre-existing alcoholic steatohepatitis. Unlike HCC, however, ICC frequently occurs in the non-cirrhotic liver,22 suggesting the need for further studies to examine the role of alcoholic fatty liver disease/alcoholic steatosis in ICC carcinogenesis.

The results of the present study also suggest that persons with extrahepatic biliary lithiasis, choledocholithiasis or cholecystolithiasis have an increased risk of developing ICC. Hepatolithiasis is a well established risk factor for ICC, and likely promotes ICC development through direct chronic inflammation and increased epithelial proliferation.23 Extrahepatic biliary lithiasis may increase the risk of ICC through a number of potential mechanisms, such as altered bile composition, cholestasis, chronic inflammation or confounding by metabolic conditions favoring the development of biliary lithiasis itself.

Cholecystectomy was of interest, in that it has been reported to be associated with a decreased risk of extrahepatic bile duct cancer 10 or more years after surgery. In the present study, however, cholecystectomy was not significantly associated with ICC although the odds ratio was consistent with a slightly increased, rather than decreased risk.24 Given that both choledocholithiasis and cholecystolithiasis were associated with increased risk of ICC, however, further examination of the association between ICC and cholecystectomy is clearly warranted.

In the present study, obesity and diabetes were more prevalent in the ICC cases compared to the controls, but neither overall association attained statistical significance. Diabetes was more common in the cases only in the year preceding ICC diagnosis, which could have been due to the diagnostic workup prior to ICC diagnosis. The current findings contrast with the findings of a recent US case-control study,10 and previous cohort studies from Denmark and Sweden that reported higher frequencies of ICC in patients with diabetes.25, 26, 27 The reasons for the differing results are not certain, but may be related to differences in the study designs. In the present study, the prevalence of diabetes was almost certainly underestimated. The risk factor assessment was based on hospital discharge records only and diabetes most commonly does not require hospitalization. In addition, the prevalence of obesity in Denmark was likely much lower during the time period of the present study than it is now and likely to be lower in Denmark vs. the United States. It is also possible that the relationship differed in the present study in contrast with the US study because of the inclusion of younger (<65 years) as well as older (≥65 years) patients.

In contrast to some prior reports, the majority of ICC cases in the current study were female. In line with prior reports, a substantial proportion of the cases (∼90%) did not have any of the risk factors under investigation.11 The high percentage of subjects with unidentified risk factors underscores the need for more studies that investigate the etiology of ICC.

Important strengths of this nationwide population-based study are related to its case and control definitions, as well as its data sources. The linkage of the citizens' identification number throughout the Danish health care system permitted the comprehensive identification of virtually all patients with ICC in Denmark. Further, the use of hospital discharge records rather than personal interview to determine pre-existing medical conditions likely avoided recall bias. The hospital discharge registry has identified more than 99% of all discharges from Danish hospitals.15 Furthermore, exclusion of ICC cases without histological confirmation minimized the possibility of misclassifying HCC or metastatic liver disease as ICC. In regard to the controls, the study was able to select among almost the entire citizenship of Denmark.

The study also had several potential limitations. The use of hospital discharge records to identify pre-existing conditions almost certainly excluded the identification of conditions that were asymptomatic or mild, thus not requiring hospitalization. As a result, some diagnoses (e.g. diabetes) are very likely to be underreported in the current study. In addition, medical conditions were selected based on the most specific available ICD-8 codes for individual liver and biliary tract diseases. This certainly increased the accuracy of the disease ascertainment, however, it might have resulted in further underreporting of corresponding medical conditions. Also, although the exclusion of risk factors that were within the year preceding the cancer diagnosis did not change the results, it is possible that some bias might have resulted from more intense medical scrutiny of the cases prior to cancer diagnosis. Finally, the absence of specific codes for chronic infection with hepatitis B virus, hepatic C virus and human immunodeficiency virus and of codes to distinguish Type 1 from Type 2 diabetes before 1987 prevented the current study from investigating these conditions. The previous US study reported significant associations between ICC and both HCV and HIV infections. However, the prevalence of HCV and HIV in the US study were 6 and 0.5% in the cases and 1.2 and 0.1% in controls, respectively. As the overall prevalence of these viral infections is lower in Denmark than in the US, and was lower yet during the time period of the current study (Statens Serum Institut, Copenhagen, Denmark), it is uncertain whether any relationship between the viruses and ICC would have been detected even had the codes been available. It was also not possible to investigate any association between H. pylori infection and ICC risk as detailed information on H. pylori infection status was not available.28

In summary, these results suggest that, in addition to previously described risk factors, alcoholic liver disease is a risk factor for ICC in Denmark. Diabetes may also be a risk factor, though it was only associated with ICC in the year prior to the cancer diagnosis in the current study. Future studies are needed to further explore the role of these risk factors and to investigate novel risk factors for ICC in low-risk populations.


The authors wish to thank Patricia Madigan for assistance in editing the manuscript.