In the present study, we found an increased risk of ICC in workers exposed to asbestos, irrespective of socioeconomic status, and smoking history. In addition, suggestive evidence of increased risk for ECC was found among workers exposed to asbestos. These findings suggest a putative role of asbestos in ICC pathogenesis and possibly in its increasing incidence.
In Italy, asbestos was used extensively for more than 100 years before it was banned in 1992. Because of the long lag time between exposure and disease development, we are now witnessing asbestos-related diseases such as mesothelioma and lung cancer .
Asbestos has been suspected as a risk factor for CC but never specifically investigated before the present study. In 1983, a case of bile duct cancer in a patient with asbestosis was described. At postmortem examination, after digesting tumor tissue, short asbestos bodies similar to those observed in the lung were recovered in the liver .
Among others, the Swedish Cancer-Environment Register reported an increased standardized incidence ratio (SIR) of CC in some asbestos-related occupations: wholesale building materials (SIR for men 2.3); shipbuilding and repair (SIR for women 7.3); insulation workers (SIR for men 10.6) . The sample size of the study was able to disclose these associations as the Swedish Cancer-Environment Register collected data on all employed Swedish citizens, followed-up for 19 years (1961–1979). Conversely, most studies on asbestos exposure and cancer were performed on worker cohorts too small to disclose an increased incidence of such a rare tumor.
The presence of asbestos fibers in the bile ducts can be explained by their translocation pathway: the fibers can cross the alveolar barrier after inhalation or penetrate the gastrointestinal mucosa after ingestion. The fibers then reach the interstitial environment and circulatory system through lymphatic vessels and are finally delivered to all tissues, where they may start an inflammatory (and hence possibly malignant transformation) process . The prerequisite for ICC pathogenesis is the presence of asbestos in the biliary tract, mainly in canals of Hering, bile ductules, and interlobular bile ducts that are the principal targets of carcinogenic agents. After their translocation from the circulatory system, asbestos fibers may remain trapped in the smaller bile ducts. This would explain why asbestos exposure seems to be involved only in ICC pathogenesis and probably not in ECC development, also considering that the multipotent stem cells putatively involved in carcinogenesis differ for ICC and ECC . Once the asbestos fibers reach the bile ducts, they could give rise to ICC through a chronic inflammation pathway, the same mechanism activated by established risk factors. Inflammatory conditions promote carcinogenesis by producing reactive oxygen and nitrogen species from inflammatory and epithelial cells, activating reparative tissue proliferation and creating a local environment rich in cytokines and other growth factors, ultimately resulting in DNA damage .
The present exploratory case–control analysis was initially based on a consecutive series of CC cases seen in our center and on historical controls. Due to its composition, the study population could be prone to referral bias. Thus, the different proportion of subjects occupationally exposed to asbestos among the ICC cases, and the population controls could be an artifact due to case selection. To address this possibility, we conducted a sensitivity analysis using only cases and controls from the Province of Bologna, a subset of our population in which a referral bias is less likely. Estimates obtained in the sensitivity analysis were in line with those obtained in the main analysis.
Because of the individual matching, we excluded from our analysis 53 cases of CC, mainly due to region of residence; hence our analysis was based on a small number of cases (41 ICC and 59 ECC). We decided to use an individual matching based on relatively small geographic areas (i.e., Italian Regions) to take into account the extreme variability and heterogeneity characterizing the economic structures in Italy (a different distribution of production activities is appreciable even among small adjacent areas). A matching based on a broader classification (e.g., North, Center, and South of Italy) would have prevented us excluding so many cases from the analysis. Nevertheless, we believe this kind of approach would have increased the risk of spurious findings, since the prevalence of subjects with a former occupational exposure to asbestos could be strongly influenced by geographical origin.
Another limitation of the present study is the retrospective assessment of occupational exposure to asbestos: we cannot exclude exposure misclassification. Unfortunately, we only had information on life-prevalent job titles for seven patients. In addition, since occupational histories were collected from relatives for 46 % of cases, it is likely that for these subjects, we missed some job titles held during their youth. On balance, occupational exposure to asbestos once in a lifetime could have been underestimated for cases compared to controls, for whom self-reported occupational histories were always available. However, were this the case, the association between occupational exposure to asbestos and CC would have been underestimated. The control group, mainly composed of population controls (75 %) and ophthalmic outpatients from the Province of Bologna (3 %) [19, 21], also included inpatients referred to Sant’Orsola-Malpighi University Hospital (22 %) . Among these subjects, exposure to asbestos could have been slightly overestimated; hence, our estimates could have been biased toward the null hypothesis (Berkson’s bias). A possible limitation of our study arises from the time when information was collected. Cases and controls were matched by age but, due to the study design, data on controls were collected 6–10 years before data on cases. Therefore, an underestimation of the number of controls occupationally exposed to asbestos could be possible, due to the truncation of occupational histories. However, the mean age of the cases at the time of the interview was 56 years (SD 11 years), and all data were collected after the national asbestos ban, introduced in 1992. Hence, a first exposure to asbestos is unlikely to have occurred among previously unexposed controls between the time of the interview and 2011. On balance, we believe that underestimation of exposure among controls, although possible, should not have substantially biased our estimates. Since exposure assessment was conducted blindly in relation to case/control status, other sources of differential misclassification (information bias) appear unlikely. Under these conditions, the risk of CC is more likely to have been underestimated than overestimated (bias toward the null hypothesis). Data of controls were mostly collected by self-administered questionnaires, although telephone administration of the questionnaire was tried among non-responders [19, 20]. Information for cases was always collected by telephone interviews. The etiological hypothesis was not mentioned during the interview, and job titles—which can be considered non-sensitive data—were collected using a structured questionnaire. Under these conditions, the quality of information on occupational history is unlikely to have been strongly influenced by the mode of data collection.
Due to the study design (case–control analysis performed with already collected data), information on possible confounders was poor. For the controls, we had no information on established/putative risk factors of CC such as parasitic infections, primary sclerosing cholangitis, bile duct cysts, hepatolithiasis, toxins, chemical carcinogens, inflammatory bowel disease, hepatitis C virus, hepatitis B virus, cirrhosis, diabetes, obesity, alcohol intake, and host genetic polymorphisms [1, 2]. Our analysis took into account possible confounding by cigarette smoking, roughly classified as never/ever smoker. We also made at least partial adjustment for socioeconomic status, classified according to the three classes version of The National Statistics Socio-economic Classification. Application of the more informative classification in five or eight classes was not feasible due to the small study population .