The “Gene-environment interactions in lymphoma etiology” (ItGxE) multicentre case-control study took place in 2011-2017 in six Italian centres, namely Perugia, Florence, Novara, Verona, Cagliari and Nuoro in Sardinia, and Bari and Taranto in Apulia. The study protocol included a questionnaire interview to the incident cases of lymphoma (any subtype) and to 2:1 age and gender frequency matched controls. In most centres, controls were hospitalized, cancer free patients from surgery wards, eye care departments, or hematology outpatients or patients suffering from trauma injuries, gastrointestinal disorders, cardiovascular disorders. Diagnoses ineligible for selection as hospital controls included malignant neoplasms (any), AIDS, autoimmune diseases, allergic diseases, viral hepatitis, organ transplants, and pre-neoplastic hematologic diseases (monoclonal gammopathy of undetermined significance, bone marrow aplasia, myelodisplastic syndrome).
In the Cagliari and Nuoro centers, controls were a random sample of the general population 2:1 frequency matched to the incident case by 5-year age and gender groups. The overall refusal rate was 7.4% among the cases and 38.4% among the controls; it was 41.1% among the population controls, and 37.1% among the hospital controls. Overall, 867 cases (500 males and 367 females) and 774 controls (428 males and 346 females) participated in the study. Trained interviewers conducted in person interviews at the hospital or at the residence of study subjects, using a modified version of the EpiLymph questionnaire [10], designed to gather information on a number of variables, including socio-demographic data, and a lifetime occupational history. For each job lasting one year or more, specific questions asked for a short description of the employer’s trade, the daily tasks, the machines and tools used, and a self-report about exposures of interest for the study. A set of 14 additional job modules, including one dedicated to gardeners and farmers, addressed further details on exposures deemed of interest based on previous epidemiological findings. The detailed information acquired through the specific job module for gardeners and farmers included type of crops, the size of each crop field, type of phytopathology treated, type of pesticides used, days/year of treatment, spraying tools, use of personal protective equipment, and post-treatment re-entry in the fields before the expiry of the respect time. Based on the above-described information, and with the support of a “crop-exposure” matrix [11], occupational physicians and industrial hygienists, with expertise in the retrospective assessment of agricultural exposures, assessed exposure to glyphosate using the following semi-quantitative indicators:
- confidence, representing the degree of certainty about whether a given study subject had actually been exposed. Three increasing level of confidence were defined depending on 1. a summary evaluation of the probability of exposure (0 = unexposed; 1= possible, but not probable; 2= probable; and 3= certain); and 2. the proportion of exposed among workers performing the same tasks (1 = ≤40%; 2 = 40-90%; 3 = ≥90%), in relation to availability of alternate herbicides, and/or local market data;
- intensity of exposure, based on the exposure circumstances (personal preparation of the pesticide mix, use of a shoulder pump or a tractor with or without a cabin, size of the surface to be treated, re-entry after treatment) and the use of personal protective equipment. Semi- quantitative exposure estimates were obtained from the publicly available EUROPOEM spreadsheet [12] (https://english.ctgb.nl/documents/assessment-framework-ppp/2016/10/27/calculation-model-europoem-ii), and classified in a four-step scale (0=unexposed; 1=low; 2=medium; 3=high). Intensity of exposure at the individual level was the highest along the work history of the study subject.
- frequency of exposure, in terms of days/year of use of the herbicide, as reported in the questionnaire and/or estimated based on the type of treatment whether curative or preventive, and the size of the surface to be treated (low frequency = ≤5 days/year; medium frequency = 5-10 days/year; high frequency = ≥11 days/year).
A score of cumulative exposure to glyphosate was then calculated as it follows [13]:
Ci= ∑ (yi X fj/3)xj
where,
Ci= score of cumulative exposure for the ith study subject;
j= jth job entry on the work history of the ith study subject;
yj= duration of exposure (in years) of the jth job entry;
xj= level of exposure intensity in the jth job entry;
fj= level of frequency of exposure in the jth job entry
Duration of exposure was approximately calculated from 1974, year of introduction of glyphosate in the market, onwards.
Lymphoma is a complex array of different neoplastic diseases that develop from the lymphatic tissue, which classification changed multiple times in the last 5 decades. Non-Hodgkin lymphoma (NHL), includes all lymphoma subtypes, independent on whether originating from B or T lymphocytes, and it excludes chronic lymphocytic leukemia (CLL) and multiple myeloma (MM). Although clinically obsolete, the NHL definition keeps being used in clinical settings to classify patients whose pathological diagnosis and immuno-histochemistry is not available yet, and by epidemiologists to preserve the possibility of making comparisons with past results. We classified lymphoma according to the 2008 update of the WHO classification of lymphoma [14], which relies on morphology, immunophenotype, molecular biology, genetics, and clinical presentation and course of the disease, and it includes all the lymphoma subtypes originating from B-lymphocytes, including CLL and MM, among the group of B-cell lymphoma (BCL), separated from T-cell lymphomas, and Hodgkin lymphoma (HL).
Statistical methods
We used unconditional logistic regression models to assess risk of lymphoma (all subtypes), the NHL and BCL groups, HL, and the major BCL subtypes, including diffuse large B cell lymphoma (DLBCL), chronic lymphocytic leukaemia (CLL), follicular lymphoma (FL), and multiple myeloma (MM), associated with ever exposure (including all categories of confidence, intensity, frequency and duration of exposure), and with categories of confidence, duration, intensity, and frequency of exposure to glyphosate, as well as with cumulative exposure to glyphosate. The following covariates were included in the regression models: age (continuous), gender, study center, and education. Education level was used as a surrogate for social class-related risk factors, and it was categorized as primary school, middle school, or vocational studies (8 years), up to high school graduation (9-12 years), and academic and university studies up to achieving degree ( 13 years).
The measure of association was the Odds Ratio (OR) and its 95% confidence interval (95% CI). We tested linear trends by the exposure metrics with the Wald test for trend (β/seβ), after continuous transformation of all the categorical variables in the regression model. We also applied the Fisher test for combined probabilities to calculate the chance probability of observing a positive trend with four different metrics bearing upon the same overall hypothesis, namely confidence, duration, frequency, and intensity of exposure, assumed as reciprocally independent [15]. We used the Cochran’s Q test to detect heterogeneity in risk across lymphoma subtypes. All the analyses were conducted with SPSS® version 20.0.
Local Ethics Committees approved the study protocol in all the participating centers. Informed consent was obtained from each participant.