[en] Retrospective dosimetry in large cohorts of medically exposed patients is essential to investigate the relationship between the dose delivered and the corresponding risk of cancer incidence over the lifetime. The estimation of doses in such studies is characterised by the poor availability and large diversity of data for the patients geometry, treatment modalities, and anatomical sites of interest. We have developed, at the Institut Gustave Roussy (IGR), two theoretical models (ICTA) and (Dos EG), flexible enough to allow their use in such diverse situations and also robust enough to allow further developments if needed. Both models are well designed to estimate the absorbed doses at numerous sites (151 anatomical points) for every patient in the cohort, and are easily handled. To simulate the patient at the time of treatment, both models use the geometrical parameters of the auxological tables (representing the mean French population). The ICTA model has been adapted to the technical characteristics of the applicators (Ra-226, Sr-90/Y-90, P-32, Y-90) that were used at the IGR for the treatment of skin hemangiomas, and has been used in a European study to evaluate dose-response relationships for radiation induced cancer. The Dos EG model has been adapted to actual treatment techniques used for external radiotherapy in different treatment centres, including various beam qualities from different machines (Cobalt, linear accelerators, betatrons, and orthovoltage tubes). It has been used to study the risk of second cancer following radiotherapy for a solid cancer in a cohort of 4400 children, and following a breast cancer in a cohort of 800 women. (author)