Background Consumption of alcohol and tobacco strongly increases risk of cancer of the tongue, mouth, pharynx, larynx, and oesophagus, and are established risk factors also for cancer of the liver, colon, and rectum. It is well documented that these habits are unequally distributed among occupational groups. Most occupational cohort studies do not have information on these potentially important confounders, and may therefore be prone to bias.

Aim The aim of the study was to calculate standardized incidence ratios (SIRs) adjusted for alcohol and tobacco by occupation, and to compare to the unadjusted SIRs.

Material and methods The study is based on the Nordic Occupational Cancer (NOCCA) database.

We used confirmatory factor analysis models where the unobserved pattern of alcohol and tobacco consumption were considered a latent common factor, and the potential occupational variation on each cancer type latent site specific factors. Results were used to compute adjusted expected numbers of cancer from the reference rates and to calculate adjusted SIRs for the relevant cancer sites for each occupation.

Results Changes of risk estimates from significantly high to significantly low and vice versa were seen. Among Nordic farmers, unadjusted SIRs for cancer of the mouth and oesophagus were 0.56 (95% confidence interval [CI] 0.51–0.61) and 0.67 (CI 0.63–0.70), respectively. After adjustment, estimates changed to 1.10 (CI 1.01–1.21) and 1.16 (CI 1.10–1.22). Unadjusted SIR for pharynx cancer among wood workers was 0.83 (CI 0.75–0.91), adjusted SIR was 1.14 (CI 1.03–1.25). For larynx cancer, results in the opposite direction were seen: unadjusted SIR for economically inactive was 1.38 (CI 1.31–1.46) while the adjusted SIR was 0.91 (CI 0.86–0.96).

Conclusion Adjustment for the latent indicators of alcohol and tobacco consumption changed risk estimates for several occupations, and may guide in the identification of true risk factors and preventive strategies.