br Fig Forest plot of associations on
Fig. 1. Forest plot of associations on alkylphenolic compounds by scenario and type of compound.
*Ajusted for age, region, education level, BMI, smoking, alcohol consumption, occupational shift, exposure to pesticides, exposure to solvents, hormonal contra-ception, postmenopausal hormone therapy, menopausal status and parity.
**Ajusted for age, region, education level, BMI, smoking, alcohol consumption, occupational shift, exposure to Tirapazamine and exposure to solvents.
2.3. Statistical analyses
The distribution of potential risk factors between cases and controls was compared using the Pearson's chi-squared test. Duration, age at first exposure, time since first exposure and time since last exposure to alkylphenolic compounds were calculated based on the years at start and stop reported for each job and the date of interview. These vari-ables were categorized in three groups using tertiles based on the dis-tribution among exposed controls. Multivariate unconditional logistic regression models were used to estimate odds ratios (OR) and 95% confidence intervals (95% CI) for the association between occupational exposures and cancer. The variables considered for inclusion in the multivariable models are shown in the Directed Acyclic Graphs (Supplemental Fig. 1). Basic adjusted models included age at interview (< 60, 60–69, ≥70), region of recruitment and educational level (primary or less, secondary, higher). Final models, in addition to the basic adjustment, included BMI (< 18.5, 18.5–24.9, 25–29.9 or ≥30), smoking status (never, current, former), alcohol consumption (never, current, past), occupational shift (day, permanent night, rotating night, other rotating shifts), ever previous occupational exposure to solvents and/or to pesticides, and family history of cancer (breast or prostate, accordingly). For breast cancer, multivariate models were further ad-justed for hormonal contraceptives use (ever, never), parity (nulli-parous, 1 child, 2 children, ≥3 children), menopause status (pre-menopause, postmenopause) and postmenopausal hormone therapy use (ever, never), given that these variables are clearly related with breast cancer risk (Risbridger et al., 2010). For all variables, missing data was ≤10% of subjects, except for alcohol consumption which was 15%. Missing values were introduced in models as independent categories. To test for linear trend, ordinal variables were treated as continuous using midpoints in the categories as category values. Multiplicative interactions were tested by means of likelihood ratio tests comparing models with and without interactions. Several sensitivity analyses were performed. Given that exposure to solvents has been previously asso-ciated to these cancers and solvents and alkylphenolic compounds may be correlated, we excluded participants who reported occupational exposure to solvent compounds to ensure that occupational exposure to these chemicals was not driving associations between exposure to al-kylphenols and cancer risk. Also, we classified those who reported oc-casional and low occupational exposure to alkylphenolic compounds as non-exposed, and thus we considered only frequent and high exposures, as a measure of high probability of exposure. We performed analyses on sub-phenotypes of breast and prostate cancer (hormone receptors status and Gleason score, respectively). Finally, we further explored adjusting our models for other measures of socioeconomic status (SES) than educational level, such as self-reported maternal and paternal SES. All analyses were conducted using Stata version 14.0 and the forest plot (Fig. 1) was graphed with R 3.4.1.
The MCC-Spain Study followed the national and international di-rectives on ethics and data protection (declaration of Helsinki and Spanish law on confidentiality of data, Organic Law on Data Protection 15/1999-LOPD). All eligible subjects signed an informed consent form for participation. Study protocol was approved by the Ethical and Research Committees of each participating center.
3.1. Demographic features of participants
Baseline characteristics of participants are described in Table 1. Compared to female controls, female cases were more likely to be younger, to have a higher BMI, to be non-smokers and never drinkers, to have been occupationally exposed to solvents, as well as to have a
family history of breast cancer, and have menarche at a younger age. Male cases were more likely to have a lower educational level, to have been occupational exposed to pesticides and to report a family history of prostate cancer, than male controls (Table 1). Female controls exposed to alkylphenols were more likely to be older than the unexposed, to live in the Barcelona region, to have a lower educational level, to have a higher BMI, to be non-smoker and never drinkers, to have been occupationally exposed to solvents or pesticides, to be parous, to have never used hormonal contraceptives, to be younger at first delivery and to be postmenopausal at the moment of the interview. Compared to the unexposed, exposed male controls were more likely to be younger, to have a higher educational level, to live in the Barcelona region, to report occupational exposure to solvents and/ or to pesticides and to have worked in permanent night shifts (Table 2).