Primary liver cancer and occupation in men: A case-control study in a high-incidence area in northern Italy



The objective of our study was to evaluate the association between occupation and risk of liver cancer. A hospital-based case-control study was carried out during 1997–1999 in the Province of Brescia, a highly industrialized area in Northern Italy with a high incidence of this neoplasm. The cases were 144 male patients with incident liver cancer (96% hepatocellular carcinoma). Controls were 283 male patients, matched to cases on age (±5 years), period and hospital of admission. Information on lifetime occupational history and alcohol consumption was obtained via interview. Specific occupational exposures to pesticides, solvents and other suspected hepatocarcinogens were evaluated. A blood sample was collected to detect hepatitis B and C infections. Odds ratios (OR) of occupational exposure and 95% confidence intervals (CI), adjusted for age, residence, education, heavy alcohol intake, hepatitis B surface antigen and hepatitis C virus antibodies positivity were computed. A statistically significant increased OR was observed for employment in repair of motor vehicles (OR 3.7; 95% CI 1.1–12.3; 9 exposed cases, 10 exposed controls). Increased ORs, although not statistically significant, were found for field-crop farm workers, food and beverage processors, blacksmiths and machine-tool operators, electrical fitters, clerical workers, manufacture of industrial machinery and personal and household services. A slightly increased OR was noted in workers exposed to toluene and xylene (OR 1.4; 95% CI 0.7–3.0, 23 cases, 36 controls); the OR was 2.8 (95% CI 1.0–7.6, 11 cases, 12 controls) for 20 or more years of exposure and 2.0 (95% CI 0.9–4.1, 21 cases, 28 controls) for 30 or more years of time since first exposure. The increase in OR seemed to be independent from that of alcohol or viral infections. Our study showed that the role of occupational exposures in liver carcinogenesis is limited. However, prolonged exposure to organic solvents such as toluene and xylene may represent a risk factor for liver cancer. © 2001 Wiley-Liss, Inc.


CI, confidence intervals; ELISA, enzyme-linked immunosorbent assay; HBsAg: hepatitis B surface antigen; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HCVAb, hepatitis C virus antibodies; IARC, International Agency for Research on Cancer; ICD, International Classification of Disease; ILO, International Labour Office; ISCO, International Standard Classification of Occupation; ISIC, International Standard Industrial Classification; OR, odds ratio; PLC, primary liver cancer; RIBA, radio-immunoblot assay; SD, standard deviation; SMR, standardized mortality ratio; TSFE, time since first employment.

Primary liver cancer (PLC) is a frequent malignancy in many regions of the world. The estimated worldwide incidence of this neoplasm is about 400,000 cases per year. The geographic areas at higher risk are Eastern Asia and sub-Saharan Africa, with age-adjusted incidence rates above 30 per 100, 000 per year.1, 2 In Italy, the mortality rate is 6–30 per 100,000 per year, higher than in many Western and Nordic European countries; the highest incidence rates have been shown in males in northeast areas of the country.3 The province of Brescia ranked among the highest in the country with an incidence rate of 35 per 100,000 in males.4

Environmental factors have a major role in the etiology of hepatocellular carcinoma (HCC), which is the most common type of liver cancer in Western countries.5 Hepatitis B virus (HBV), hepatitis C virus (HCV) and alcohol drinking are the major known risk factors for PLC in developed countries.6–9

With the exception of the well-established association between occupational exposure to vinyl chloride monomer and liver angiosarcoma,10 the relationship between occupation and PLC has not been extensively studied. So far, only 2 epidemiologic studies aimed at investigating occupational exposures as risk factors for PLC took also into account alcohol consumption and viral infections.11, 12 In fact, the literature reports relatively few relevant studies and most of them were conducted before valid biologic markers for HCV infection became available for large-scale investigations.13–17

The aim of our study was to evaluate the association between occupation and PLC. This research was an extension of previous studies carried out in the same area to investigate the association between HCC and infection by HBV, HCV and other newly discovered viruses, alcohol drinking and family history of liver cancer, results of which have been reported previously.18–23


A hospital-based case-control study began in spring 1997 and lasted 2 years. The research included subjects born in Italy, aged up to 75 and resident in the province of Brescia, Northern Italy. Since HCC is 3 times more common among men than women in this area4 and occupational exposures are more prevalent in men, the analysis was restricted to men. We examined 149 male liver cancer incident cases, who were admitted to the 2 main hospitals in the province. PLC diagnosis (ICD-IX: 155.0 and 155.1) was assessed by histology, cytology, α feto protein levels ≥400 ng/mL, sonography or computerized tomography. Controls were 302 men admitted to various departments of the same hospitals. Subjects affected by malignant tumors or liver diseases were excluded. Controls were frequency matched on cases by age (±5 years) and by date and hospital of admission.

Nineteen controls who refused the interview and 5 cases who gave inadequate interviews were excluded from analysis; thus, the final population consisted of 144 cases and 283 controls. Among these subjects, 56 cases and 93 controls had already been included in a previous study regarding nonoccupational risk factors.22

The size of our study was based on a calculation of statistical power, which showed a requirement of 120 cases and 240 controls to estimate an OR of 2 or more for occupations and industrial activities with 10% exposure proportion, with an α error of 0.05 (2-sided test) and a statistical power of 0.80.

Our study was approved by the Ethical Committee of the major hospital involved in the research. A written informed consent describing aims, methods and indicating persons responsible for the study was obtained by each patient.

Cases and controls were interviewed face to face during hospital admission by 2 residents in occupational medicine, who were aware of the case-control status of study subjects. A structured questionnaire included questions on socio-demographic aspects, education, lifetime tobacco and alcohol consumption, leisure-time activities and occupational history. Total alcohol intake was assessed according to the average ethanol content of wine (12% by volume), beer (5%) and spirits (40%); heavy alcohol intake was defined as an intake of more than 60 g ethanol per day for at least 5 years. Lifetime occupational history was collected for each job lasting at least 1 year. The information regarded job title, plant activity, type of production and exposures to chemicals and physical agents, with detailed descriptions of workplaces and job tasks as well as the use of personal protective devices and hygienic behaviours in workplaces. The average duration of the interview was about 1 hr. Job titles and plant activities were coded according to the International Standard Classification of Occupation24 and the International Standard Industrial Classification of all economic activities.25

Codes were attributed by an occupational physician unaware of the case-control status of the subject. A random 10% sample of questionnaires was independently recoded by another expert; of 450 codes, less than 3% showed a major disagreement (i.e., 1st or 2nd digit), which was subsequently discussed and resolved.

For each specific job title, it was assessed whether the subject was specifically exposed to the following substances, which are known or suspected liver carcinogens and chosen a priori on the basis of the literature: pesticides (particularly organophosphorous, organochlorine, arsenicals, herbicides), chlorinated solvents (particularly trichloroethylene, tetrachloroethylene, carbon tetrachloride, methylene dichloride), nonchlorinated solvents (xylene and toluene), polychlorinated biphenyls, polybrominated biphenyls, vinyl chloride monomer, aflatoxins. Following an approach developed in other studies,26, 27 these specific exposures were classified according to 4 exposure indices, such as reliability (i.e., the degree of confidence that the exposure had actually occurred and classified as certain, probable, possible), route (respiratory, dermal, both), intensity (low, medium, high) and frequency (low [1–5% of normal working time], medium [5–30%], high [more than 30%]). An occupational physician, blind as to case-control status, converted job titles into specific exposures on the basis of the detailed occupational history, his specific experience about local industrial milieu, information provided by local health and safety consultants, company records on risk assessments and scientific and technical literature and also taking into account temporal variation of the exposure pattern. An exposure coding sheet was prepared, listing the above mentioned specific exposures, which were systematically considered for each job title, together with the 4 exposure indices.

Leisure-time activities entailing chemical exposure were also coded according to the same criteria. A random 5% sample of questionnaires attributing specific exposures was coded independently by another occupational physician for quality control. Only few and minor disagreements were found and subsequently resolved after discussion.

A 10 mL blood sample was taken from all subjects and serum was stored at −80°C until analysis. The presence of the hepatitis B surface antigen (HBsAg) and of HCV antibodies (HCVAb) was assessed using commercial enzyme-linked immunosorbent assays (ELISA; Abbott Labs, North Chicago, IL). HCVAb-positive sera were tested for confirmation by an immunoblot assay (RIBA; Ortho, Raritan, NJ).

ORs of liver cancer and 95% CIs were computed by unconditional logistic regression analysis using the maximum likelihood method. In exploratory analyses, both 90% and 95% CI are commonly used; we preferred the latter in order to compare our results with other literature findings on liver cancer and occupation. Potential confounding was accounted for by computing OR adjusted for age, area of residence (city of Brescia vs. rest of province), education, alcohol consumption, smoking habits and HCVAb and HBsAg positivity. Only adjusted ORs are shown in the tables. Interaction between alcohol intake, viral infections and occupational exposures was also assessed.

A total of 343 occupations and 144 industrial categories were analyzed. The results are reported in the tables for some selected occupations and industries on the basis of relevance of the occupation or industry in the province of Brescia (such as metalworking, construction industry, transports, textile industry, agriculture, various services), a priori indications by the scientific literature and if there were at least 7 exposed subjects of whom there were at least 1 case and 1 control. Full results are available on request.

Analyses by time since first employment (TSFE) and analysis by duration of employment were also performed. Two-sided statistical tests were performed using an alpha value of 0.05 for refusing the null hypothesis.

All the analyses were performed using the BMDP (Dynamic version) program for a personal computer.


Distribution of cases and controls according to socio-demographic variables is shown in Table I. Cases had a lower education level than controls. The mean age at interview was 63.1 years among cases and 62.6 in controls (p > 0.1). The mean number of jobs was similar in cases and controls.

Table I. Sociodemographic Characteristics of Cases and Controls
  • 1

    p < 0.0001.

  • 2

    p > 0.1.

  • 3

    p < 0.01.

  • 4

    p > 0.1.

 City of Brescia68(47.2)195(68.9)
Age (years)2
 < 509(6.2)23(8.1)
Education (years)3
 ≥ 913(9.0)58(20.5)
Mean no. of jobs (S.D.)43.68(±1.76)3.53(±2.1)

The distribution of cases according to the diagnostic procedures is reported in Table II. As expected, the histology showed a high proportion of HCC; no angiosarcomas were detected. Controls suffered from a variety of diseases of the genitourinary (22.6% of controls), digestive (21.9), circulatory (21.5) and respiratory (5.6%) systems, dermatologic (5.3) and metabolic (4.9%) disorders, traumas (4.2%), ill-defined symptoms (12.4) and other conditions (1.4). Table III shows the selected ORs according to job title. No statistically significant increased risks were detected. However, nonsignificant excess risks were observed for motor vehicle mechanics, field-crop farm workers, food and beverage processors, electrical fitters, machine-tool operators and clerical workers. Among industrial activities (Table IV), a statistical significant risk was found only for repair of motor vehicles; increased risks were found for electricity gas and steam, manufacture of furniture and of special industrial machinery, transport (and in particular railways) and personal services.

Table II. PLC Cases: Diagnostic Protocol
Biopsy (histotype)98(68.1)
 Hepatocellular carcinoma94(95.9)
α-fetoprotein ≥ 400 ng/mL8(5.6)
Sonography or computerized tomography34(23.6)
Table III. Odds Ratios of Liver Cancer According to Job Title
Job category (ISCO code)Cases/controls exposedOR1 (95% CI)
  • 1

    OR estimated by logistic regression analysis adjusted for age, residence, education, HBsAg and HCVAb positivity and heavy alcohol consumption.

Clerical and related workers (3)21/402.0 (0.9–4.2)
Agricultural, animal husbandry and forestry workers, fishermen and hunters (6)41/650.8 (0.5–1.5)
Agricultural and animal husbandry workers (62)32/371.3 (0.7–2.5)
Field-crop and vegetable farm workers (622)10/81.8 (0.6–5.6)
Metal processors (72)17/370.5 (0.2–1.1)
Food and beverage processors (77)16/171.6 (0.6–3.8)
Cabinetmakers and related woodworkers (81)9/121.2 (0.4–3.5)
Blacksmiths, toolmakers and machine-tool operators (83)40/621.5 (0.8–2.7)
Machine-tool operators (834)31/461.5 (0.8–3.0)
Machinery fitters, machine assemblers and precision instrument makers (except electrical) (84)25/451.4 (0.7–2.8)
Motor vehicle mechanics (843)7/82.7 (0.8–11.7)
Electrical fitters and related electrical and electronics workers (85)9/111.7 (0.5–5.6)
Painters (93)7/100.6 (0.1–2.6)
Bricklayers, carpenters and other construction workers (95)32/580.9 (0.5–1.6)
Material-handling and related equipment operators, dockers and freight handlers (97)23/351.0 (0.5–2.2)
Warehouse porter (97145)12/150.9 (0.3–2.9)
Transport equipment operators (98)18/291.0 (0.5–2.3)
Motor vehicle drivers (985)14/260.8 (0.3–2.0)
Table IV. Odds Ratios of Liver Cancer According to Industrial Activity
Industrial activity (ISIC code)Cases/controls exposedOR1 (95% CI)
  • 1

    OR estimated by logistic regression analysis adjusted for age, residence, education, HBsAg and HCVAb positivity, heavy alcohol consumption.

Agricultural and livestock production (1110)23/450.8 (0.4–1.7)
Agricultural services (1120)18/201.2 (0.5–2.8)
Manufacture of food, beverages and tobacco (31)13/240.8 (0.3–2.1)
Manufacture of furniture and fixtures, except primarily of metal (3320)9/102.0 (0.6–6.9)
Manufacture of chemicals and of chemical, petroleum, coal, rubber and plastic products (35)7/230.8 (0.2–2.5)
Iron and steel basic industries (3710)18/341.0 (0.4–2.2)
Manufacture of fabricated metal products, except machinery and equipment (381)23/440.6 (0.3–1.3)
Manufacture of machinery except electrical (382)33/541.4 (0.7–2.7)
Manufacture of metal and woodworking machinery (3823)8/171.8 (0.6–5.3)
Manufacture of special industrial machinery and equipment except metal and woodworking machinery (3824)10/122.4 (0.8–7.2)
Machinery and equipment except electrical not elsewhere classified (3829)19/281.1 (0.5–2.6)
Electricity, gas and steam (41)4/42.6 (0.4–15.7)
Construction (5000)40/760.8 (0.4–1.4)
Restaurants and hotels (63)11/151.5 (0.6–4.2)
Transport, storage and communication (7)18/251.7 (0.7–3.9)
Railway transportation of goods not elsewhere classified (7170)6/53.0 (0.6–14.4)
Personal and household services (95)16/242.0 (0.8–4.7)
Repair of motor vehicles and motorcycles (9513)9/103.7 (1.1–12.3)

The analysis according to TSFE could not be performed for many occupations and industries, since most workers were in the category with 30 or more years of TSFE (data not shown). For example, in the field-crop and vegetable farm workers category (ISCO code 622), an OR of 1.2 (95% CI 0.1–21.5) was found in subjects with less than 30 years of TSFE (2 cases, 1 control), whereas the risk was 1.9 (0.5–6.5) for subjects with more than 30 years (8 cases, 7 controls); for motor vehicle mechanics, there was only 1 control exposed in the less than 30 years category, whereas the risk was 2.9 (0.7–12.1) in the other category (7 cases and 7 controls).

The analyses by duration of employment were restricted to the following occupations, with sufficient numbers of exposed cases and controls: clerical workers, agricultural workers, food and beverage processors, machine-tool operators, manufacture of various machinery and personal services. The logistic regression analysis did not show systematically higher liver cancer risk for subjects with more than 20 years of duration of employment compared to those with 1–19 years (data not shown).

The analysis of specific occupational exposures (Table V) was performed only for nonchlorinated organic solvents (i.e., mainly toluene and xylene, since the use of other nonchlorinated organic solvents are relatively uncommon in the area of Brescia), owing to the fact that only 2 cases and 12 controls were exposed to pesticides (especially herbicides) and 3 cases and 12 controls were exposed to chlorinated solvents (mainly perchloroethylene and trichloroethylene). A slightly increased risk was found for overall exposure to nonchlorinated solvents. No consistent pattern was observed when the ORs were computed according to type, level and frequency of exposure to solvents, whereas a higher risk (not significant) was noted in subjects with “certain” exposure compared to those with probable exposure. A borderline statistically significant increased risk was noted in subjects with 20 or more years of “certain” exposure. When duration of exposure was evaluated according to level and frequency of exposure, we did not observe any linear trend in the cases/controls ratio by categories of duration for either level or frequency of exposure. However, the small number of subjects in each category of the combination of the 2 variables limited the value of this analysis.

Table V. Assessment of Occupational Exposure to Nonchlorinated Organic Solvents (Toluene and Xylene)
VariableCases/controlsOR1 (95% CI)
  • 1

    OR estimated by logistic regression analysis adjusted for age, residence, education, HBsAg and HCVAb positivity and heavy alcohol consumption. Reference category: no overall exposure for each analysis.;2>

Overall exposure
 Yes23/361.4 (0.7–3.0)
 Probable5/160.8 (0.2–3.1)
 Certain18/201.9 (0.7–4.6)
Type of exposure
 Respiratory1/60.5 (0.1–4.6)
 Dermal3/33.6 (0.6–22)
 Respiratory and dermal19/271.4 (0.6–3.3)
Level of exposure
 Low8/171.9 (0.6–2.3)
 Medium8/101.2 (0.3–4.3)
 High7/91.0 (0.2–4.1)
Frequency of exposure
 Low9/181.4 (0.5–3.9)
 Medium7/81.7 (0.4–6.4)
 High7/101.1 (0.3–4.8)
Duration of exposure
 Solvents1–1911/171.2 (0.4–3.5)
  20+12/191.6 (0.6–4.3)
 Solvents (exposure “certain”) 1–197/81.5 (0.4–5.5)
  20+11/122.8 (1.0–7.6)

Among the subjects belonging to the category “repair of motor vehicles” (9 cases and 10 controls), there were 5 cases and 6 controls exposed to nonchlorinated solvents (OR 3.6, 95% CI 0.8–16).

Analysis according to TSFE for such solvents showed an OR of 0.5 (95% CI 0.4–3.1, 2 cases, 8 controls) in the less than 30 years category and 2.0 (0.9–4.1, 21 cases, 28 controls) in those with 30 or more years of TSFE.

When the nonoccupational risk factors for PLC were considered, all but 6 cases were either heavy alcohol drinkers or positive for HBV or HCV infection. As expected, strong associations were found between these factors and liver cancer.

Tables VI, VII and VIII show the results of the interaction analysis between selected occupational exposures and alcohol intake, HBV and HCV infection. No systematic effect modification could be observed in the ORs. However, among workers without either history of heavy alcohol intake or HBV or HCV positivity, a higher risk of PLC was noted in motor vehicle mechanics (statistically significant), in blacksmith-toolmakers and in subjects specifically exposed to nonchlorinated organic solvents.

Table VI. Odds Ratios of Liver Cancer and Job Title According to the History of Heavy Alcohol Intake
Job category (ISCO code)Ever employedAlcohol intake
≤ 60 g/day> 60 g/day
Cases/controlsOR1 (95% CI)Cases/controlsOR1 (95% CI)
  • 1

    OR estimated by logistic regression analysis adjusted for age, residence, education and HBsAg or HCVAb positivity.

Agricultural and animal husbandry workers (62)No36/157176/891
Yes7/161.0 (0.2–4.1)25/211.5 (0.7–3.1)
Food and beverage processors (77)No40/166188/1001
Yes3/71.6 (0.2–12.8)13/101.5 (0.6–4.1)
Machine-tool operators (834)No34/147179/901
Yes9/261.3 (0.4–4.5)22/201.4 (0.6–3.1)
Blacksmith, toolmakers and machine-tool operators not elsewhere classified (839)No40/165194/1041
Yes3/81.6 (0.2–11.1)7/60.1 (0.0–0.3)
Motor vehicle mechanics (843)No40/168197/1071
Yes3/59.2 (1.1–77.8)4/32.4 (0.4–14.2)
Table VII. Odds Ratios of Liver Cancer and Job Title According to the Presence of Viral Hepatitis Infection (HBV, HCV)
Job category (ISCO code)Ever employedHBV and HCV positivity
Both negativeAt least one positive
Cases/controlsOR1 (95% CI)Cases/controlsOR1 (95% CI)
  • 1

    OR estimated by logistic regression analysis adjusted for age, residence, education and alcohol intake.

Agricultural and animal husbandry workers (62)No45/221167/251
Yes21/302 (0.9–4.2)11/70.4 (0.1–1.5)
Food and beverage processors (77)No8/16170/311
Yes58/2351.3 (0.5–3.7)8/13.1 (0.3–32.1)
Machine-tool operators (834)No54/210159/271
Yes12/411.2 (0.5–2.8)19/52.1 (0.6–7.3)
Blacksmith, toolmakers and machine-tool operators not elsewhere classified (839)No60/239174/301
Yes6/122.0 (0.5–7.1)4/21.1 (0.1–8.5)
Motor vehicle mechanics (843)No63/243174/32
Yes3/83.1 (0.6–16.9)4/0
Table VIII. Odds Ratios of Liver Cancer and Nonchlorinated Organic Solvents Exposure, According to the Presence of Alcohol Intake and Viral Hepatitis Infection (HBV, HCV)
Exposure to solventsAlcohol intakeHBV or HCV infection
≤ 60 g/day> 60 g/dayBoth negativeHBV or HCV positivity
Cases/controlsOR1 (95% CI)Cases/controlsOR1 (95% CI)Cases/controlsOR2 (95% CI)Cases/controlsOR2 (95% CI)
  • 1

    OR estimated by logistic regression analysis adjusted for age, residence, education and HBV and HCV infections.

  • 2

    OR estimated by logistic regression analysis adjusted for age, residence, education and alcohol intake.

Yes20/161.81 (0.8–4.3)3/200.8 (0.1–4.7)14/22.7 (0.5–15.2)9/341.2 (0.4–3.0)


Previous reports on the role of nonoccupational risk factors indicated that most HCC cases in the area of Brescia, Northern Italy, were attributable to HCV, HBV and alcohol consumption.9, 19, 22 However, the area under study is highly industrialized, with extensive metal and mechanical industries, construction, manufacture of textiles, agriculture, various services and with exposure to several xenobiotics. Therefore, a possible role of occupation on PLC etiology could be evaluated.

Our study showed a significantly increased risk for motor vehicle mechanics and repair, though no trend with duration of exposure was evident, mostly because of the small number of exposed cases and controls. The possibility of this being chance result may not be ruled out due to the multiple comparisons. However, in these occupations exposure is possible to known or suspected human carcinogens, such as polycyclic aromatic hydrocarbons, chlorinated and nonchlorinated solvents, engine exhausts, mineral oils and paints.28–31 Moreover, other studies reported a nonsignificant OR of 1.6 for liver cancer in mechanics and repairers14 and an OR of 5.4 (95% CI 2.2–13.6) for auto repair shops.17

A modest excess risk was also found for occupations with easy access to alcoholic beverages, such as food and beverage processors and in restaurant activities, in agreement with previous studies showing significant risk for PLC in workers employed in breweries, distilleries, restaurants, hotels, motels17 and in workers employed in eating and drinking places and as cooks or bartenders.13, 15 Overall, these might be spurious associations, possibly caused by underreporting of alcohol intake.

Previous studies found increased risks in farm laborers,11, 13, 16 which is explained by possible exposure either to pesticides or to agricultural products contaminated by aflatoxins. However, other surveys did not detect excess liver cancer risk in agriculture.14, 15, 17 We did not observe a significantly increased risk in agricultural workers, however, exposure to pesticides in our population was indeed limited.

We found an elevated liver cancer risk following occupational exposure to nonchlorinated organic solvents such as toluene and xylene. The overall risk for ever and “certain” exposed workers was not significantly increased, but a trend for duration of exposure was noted and an OR of 2.8, at the limit of statistical significance, was detected in workers with >20 years of exposure classified as “certain”.

Moreover, the risk was also elevated in workers with >30 years of TSFE, in subjects without history of heavy alcohol intake and in those without HCV or HBV infections, suggesting a specific role for nonchlorinated solvents independent of the other main risk factors for HCC. All these data point toward an effect of occupational exposure to nonchlorinated organic solvents in the etiology of PLC.

We should note that the literature reports meta-analyses in which a significant risk for liver cancer was detected in painters32 and in workers exposed to organic solvents.33 However, in many of these studies, due to concurrent exposure to various xenobiotics, solvent-specific risk estimation could not easily be made.

Chlorinated solvents such as perchloroethylene and trichloroethylene are suspected human liver carcinogens.34–37 Our data do not seem to support their role as hepatocarcinogens, even though we could not properly evaluate the role of chlorinated solvents in our study because of the small numbers of cases and controls exposed to such xenobiotics.

Among the potential limitations of our study, selection bias should be first discussed. This bias is unlikely to have occurred, since the cases included in the study were an unselected sample of all incident liver cancers arising in the province of Brescia. In fact, in a previous study on liver cancer in patients with or without cirrhosis, we verified the population basis assumption by comparing the cases included in our study with the hospital discharge data; among the total cases arising in the area, 86% were actually admitted to the 2 hospitals from which we recruited the cases.18

Controls were affected by a variety of acute and chronic nonneoplastic diseases, for which no association with the main risk factors—both nonoccupational and occupational—for liver cancer has been reported. Furthermore, almost all eligible cases and controls agreed to participate in our study, so a response-rate bias was also unlikely.

To check the presence of differential exposure misclassification due to interviewers' bias or differential reporting, duration of occupational interviews and number of jobs held, inner coherence between related questions were evaluated among cases and controls and no relevant differences were noted. Moreover, the interviewers were not blind in regard to the case-control status of the patient because of logistic reasons. However, should such a bias arise, it is likely to be against the null hypothesis, causing more jobs and industrial categories to be associated with the disease than they actually are. Our results lead, on the contrary, to the opposite direction because only a few categories were associated with PLC.

Misclassification of job titles and of expert assessments of an individual's exposure status is an important issue in retrospective epidemiologic studies, since interview data may lack precision in qualitative and quantitative definition of exposures and may have effects on the measures of relative risks and on the statistical power. The commonly suggested rating methodology that we used, which provides flexibility especially in exposure assessment and which might be particularly useful for hypothesis-generating studies like ours,38, 39 may have reduced misclassification and may have yielded sufficient information for interpretation of specific occupational exposures.

Nonetheless, a residual nondifferential misclassification cannot be eliminated and, therefore, one would expect an attenuation of OR estimates and loss of statistical power thus missing possible causal associations. Exposure-response gradient might also be attenuated.40

The statistical power of our study was low for many categories and this fact may not enable the identification of true associations and cause imprecision of the corresponding OR. However, our study had enough power to investigate an OR of 2 or more for solvent exposures and also the 95% CI did not include the null value, at least for subjects with more than 20 years of exposure to nonchlorinated solvents.

Finally, misclassification due to liver cancer misdiagnosis is unlikely since about 70% of PLC cases had a histologic/cytologic assessment and the remaining cases were diagnosed thorough validated biochemical and instrumental procedures.41 Furthermore, separate analyses of cases with and without histologic assessment showed no differences in the prevalence of HBV and HCV infections and heavy alcohol consumption as well as no differences in the socio-demographic characteristics. Also, analysis restricted to histologically confirmed cases showed an OR of 1.8 (0.8–4.1) for subjects exposed to xylene and toluene, which was similar to the OR found for the whole population.

In conclusion, our study suggested that the role of occupational risk factors in the etiology of PLC in a high-risk area from Italy is overall limited. In certain industries or occupations, such as motor vehicles repair, an increased risk might be present. Exposure to nonchlorinated organic solvents such as xylene and toluene might represent a risk factor for PLC. These results, if confirmed, should be taken into account when planning preventive measures in occupational settings.