Reproductive Factors, Exogenous Hormone Use, and Risk of B-Cell Non-Hodgkin Lymphoma in a Cohort of Women From the European Prospective Investigation Into Cancer and Nutrition

Laura Costas; Leila Lujan-Barroso; Yolanda Benavente; Naomi E. Allen; Pilar Amiano; Eva Ardanaz; Caroline Besson; Heiner Boeing; Bas Bueno-de-Mesquita; Iris Cervenka; Renée T. Fortner; Agnès Fournier; Marc Gunter; Sophia Harlid; José María Huerta; Mats Jerkeman; Karin Jirström; Rudolf Kaaks; Anna Karakatsani; Kay-Tee Khaw; Anastasia Kotanidou; Eiliv Lund; Giovanna Masala; Amalia Mattiello; Beatrice Melin; Virginia Menéndez; Neil Murphy; Alexandra Nieters; Kim Overvad; Elio Riboli; Carlotta Sacerdote; Maria-Jose Sánchez; Julie A. Schmidt; Sabina Sieri; Anne Tjønneland; Antonia Trichopoulou; Rosario Tumino; Roel Vermeulen; Elisabete Weiderpass; Silvia de Sanjosé; Antonio Agudo; Delphine Casabonne


Am J Epidemiol. 2019;188(2):274-281. 

In This Article


EPIC is an ongoing multicenter cohort study that recruited 521,324 participants between 1992 and 2000 from 23 research centers in 10 European countries. Participants were generally recruited from the general population residing in a geographic area. Exceptions were the cities of Utrecht, the Netherlands, and Florence, Italy (women participating in breast cancer screening programs); parts of the Italian and Spanish cohorts (blood donors); most of the Oxford, United Kingdom, cohort (vegetarian volunteers); and participants from France and Germany (health care insurance organizations). In France, Norway, Utrecht, and Naples, Italy, only women were enrolled.[12] At recruitment, participants signed a consent form and provided information on diet and lifestyle, and anthropometric measurements were taken. Data collection procedures were centralized as those of a single study with multiple centers. Specific questionnaires for women were used to collect information on menstrual factors, reproductive history, and use of exogenous hormones.[12] Participants with prevalent cancer (except nonmelanoma skin cancer) and those with missing follow-up information were excluded (n = 29,332). Men (n = 148,007) and persons with incomplete information on lifestyle factors (n = 527) were excluded from the present analysis.

Incident lymphoma cases were identified through population cancer registries and active follow-up (through 2015), including use of health insurance records, hospital registries, and direct contacts with participants or next of kin. Lymphoid neoplasms were initially classified according to the International Classification of Diseases for Oncology, Second Edition (ICD-O-2), and were then later recoded to the International Classification of Diseases for Oncology, Third Edition, from the World Health Organization Classification of Tumours of Haematopoietic and Lymphoid Tissues.[13] The conversion was made using an algorithm available on the Surveillance, Epidemiology, and End Results web page ( and involved a pathology expert and local expertise from participating EPIC centers. Cases with ICD-O-2 codes that could not be translated unequivocally into a lymphoid neoplasm diagnosis according to the World Health Organization classification system were categorized as "lymphoid neoplasm, unclassified" ("NOS"). The classification was further revised by participating centers using the InterLymph Pathology Working Group classification, which is based in the 2008 World Health Organization classification.[13] We refer here to "B-cell NHL," which is equivalent to "mature B-cell neoplasms" as defined by the World Health Organization and which includes multiple myeloma and chronic lymphocytic leukemia/small lymphocytic leukemia in its definition. The total number of cases included 1,849 lymphomas, of which 1,427 were B-cell NHL, 73 were T-cell NHL, 80 were Hodgkin lymphoma, and 269 were other unclassified subtypes of lymphoma. The 1,427 B-cell NHL cases were further categorized into 302 cases of diffuse large-cell lymphoma, 264 cases of follicular lymphoma, 289 cases of chronic lymphocytic leukemia/small lymphocytic lymphoma, 387 cases of multiple myeloma, and 185 cases of other subtypes of B-cell NHL. Analyses of Hodgkin lymphoma, T-cell NHL, and other unclassified subtypes of lymphoma were not performed owing to small numbers. Thus, the present analyses were based on 343,458 women and 1,427 cases of B-cell NHL.

Information on variables included in the analysis was collected at baseline using standardized questionnaires. These variables included reported age at menarche, number of full-term pregnancies, age at first birth, breastfeeding, duration of breastfeeding, history and duration of oral contraceptive use and postmenopausal hormone therapy, menopausal status, reported age at natural menopause, oophorectomy, and hysterectomy. For the hormone therapy variables, participants were asked whether they had ever used these drugs and about the timing of use, age at starting use, total duration of use, and type of formulation (estrogen alone, progestin alone, or estrogen + progestin). Self-reported baseline menopausal status was defined as menopausal (natural cessation of menses in the last 12 months or surgical menopause due to bilateral ovariectomy), perimenopausal (no longer naturally menstruating at the time of recruitment or fewer than 9 menstrual cycles in the past 12 months), and premenopausal (regular menses or at least 9 menstrual cycles in the past 12 months).

Proportional hazards modeling was used to estimate hazard ratios and 95% confidence intervals for reproductive factors and risk of B-cell NHL and its major subtypes. Age was used as the underlying time scale, and all models stratified by age at recruitment (1 year-categories) and study center and adjusted for educational level. Body mass index, physical activity, and smoking status were not included as adjustment covariates because they did not change the risk estimates by more than 10%. The proportional hazards assumption was checked using graphical methods and a goodness-of-fit test. Additional analyses were performed by means of the Wald test statistic to assess the homogeneity of the risk between lymphoma subtypes, using the SAS macro %SUBTYPE (SAS Institute, Inc., Cary, North Carolina).[14] All analyses were performed with SAS, version 9.4.