Avaluació de la taxa de detecció de càncer i del càncer d’interval en programes de detecció precoç del càncer de mama utilitzant mètodes longitudinals

Resumen

Introduction. Early detection of breast cancer (BC) has been shown to be the best strategy to reduce long-term mortality and improve prognosis. Different studies on the effectiveness of this strategy show a lower reduction in mortality and producing adverse effects. This thesis focuses on the evaluation of BC determinants in screening and its adverse effects -false positives (FP) and false negatives (FN). Objectives. This thesis aims to provide greater knowledge of the evaluation of population-based breast cancer screening programmes, specifically by 1) quantify the accumulated risk of detecting cancer in the screening for 10 years of participation and evaluating its associated factors; 2) quantify the risk of having interval cancer (IC) and evaluating its associated factors; 3) compare different methods to estimate the effect of a risk factor in the risk of presenting a BC in the participants; 4) compare the biological characteristics of IC with those of the tumours detected in the screening tests. Methods. In the first objective, we study the screening detection rate with the retrospective cohort of the RAFP project. We estimate the relationship between the detection rate and the characteristics of the mammography protocol and those of women with survival methods in discrete-time. In the second objective, we study the IC rate with the retrospective cohort of the INCA project. We estimate the relationship between the rate of IC and detection with the risk factors according to the characteristics of the protocol and those of women with a Cox model with proportional and competitive risks cause-specific. In the third objective, we compare the previous models with a multi-state model based on a Markov process. To compare them, we use the simulation and apply it to the Catalan population of the INCA project. In the fourth objective, we designed a case-control of the cancers detected in the screening and the IC of the INCA project. IC have been classified radiologically (true intervals, FN, occult tumours and minimal signs). We use a multinomial model to study the differences between them and those detected in the screening. Results. 1) The accumulated risk of detecting in the screening for 10 years is between 11.11 and 16.71 for each 1,000 participants according to the variables of the protocol. Using double reading and two projections is the strategy that detects more. 2) Having a prior FP result is the main risk factor for IC, especially for FN. The characteristics of women are risk factors regardless of the detection mode. 3) The multi-state and Cox models correctly estimate the effect of the FP on the appearance of the CM. The multi-state models allow estimating the transitions of not having cancer at the preclinical stage and from preclinical to clinical. 4) Protocol variables are associated with detection in screening and not with IC. For phenotypes, we have two groups the true intervals and minimum signs; and FN and occult tumours -similar to those detected in screening. Conclusions. A better knowledge of the adverse effects of the BC screening should allow it to be improved. IC risk factors are the same as those detected in screening. By subtype, we have a division between true intervals and minimal signs; and FN and occult tumours. The multi-state models allow to model the natural history of the BC and include participation in the screening in the model to better evaluate the effect of the variables.