Breast cancer is the most common cancer among women worldwide. In most countries, women are invited for breast cancer screening based on age. Screening programmes aim to diagnose cancer earlier and reduce surgical therapies such as mastectomy or breast cancer deaths. However, mammography can yield false-negative results and false-positive suspected breast cancer cases, resulting in unnecessary diagnostic procedures (biopsies) and therapies. In contrast to age-based screening, risk-based screening considers multiple risk factors. Breast cancer should thus be detected at least equally well or earlier, and the disadvantages of age-based screening should be reduced, for example, through less frequent mammographies in women at low risk of breast cancer. The objective of this report was to investigate the risk prediction quality of identified risk prediction models and their effectiveness in a risk-based screening programme. In addition, organisational requirements for implementing a risk-based screening programme were summarised.

Two systematic searches and additional internet searches were conducted. Based on the predefined inclusion criteria, a total of 107 studies were considered in eight systematic reviews and examined the prognostic quality of 30 risk prediction models. No randomised controlled trials were completed assessing the risk-based screening programmes' benefit-harm ratio. Concerning relevant implementation aspects, 48 references were considered, of which 21 were from the systematic search, and 27 were from additional sources.

Regarding the accuracy of the individual risk estimation of the prediction models, the studies found limited ability to predict individual breast cancer risk regardless of the model. Only one model showed good discriminatory accuracy. However, it calculated risk for a specific breast cancer type (hormone receptor status "ER-positive, HER-2-negative"), whereas all other models predict breast cancer risk regardless of hormone receptor status. Regarding the quality of the population-level risk prediction, the studies found that two empirical models, the Breast Cancer Surveillance Consortium (BCSC) model and the Rosner-Colditz model, and one genetic model, the International Breast Cancer Intervention Study (IBIS) model, had good risk prediction in the populations studied.

Before implementing a risk-based screening programme, it is necessary to define which and how many risk factors should be assessed, to what extent, and by whom, using which predictive model. Since the risk results alone have no benefit for women, risk-based recommendations on the screening interval and possible preventive measures must be decided. Risk-based recommendations require thresholds at which five-, ten-, or lifetime risk a woman falls into a high-, medium-, or low-risk group.

The successful use of a risk-based screening strategy is based on the prognostic quality of the predictive models and whether the risk-based screening recommendations and preventive interventions are effective, appropriate, accessible, feasible, and acceptable. Only the large trials currently comparing a risk-based screening strategy with conventional age-based screening will show whether this is reliably the case.