Background: PREDICT Breast is an online prognostication and treatment benefit tool to aid clinical decision making for patients with early invasive breast cancer. Since its development in 2010, the model has shown a variety of prognostic outcomes among numerous studies. Due to improvements in breast cancer survival and advancements in cancer treatments, the model might be outdated and could lead to imprecise survival predictions. The aim of this doctoral research was to update the model and enhance its model performance in order to contribute to more accurate predictions for individual patients. The first aim was to investigate and incorporate the prognostic effect of the biomarker progesterone receptor (PR) into the model. Another objective was to address the underestimation and overestimation of breast cancer mortality by amending the prognostic tool with more recent data. Methods: The prognostic effect of PR status was based on the analysis of data from 45,088 breast cancer patients of European descent from 49 studies in the Breast Cancer Association Consortium (BCAC). Cox proportional hazards models were used to obtain estimates of the relative hazard for breast cancer-specific mortality associated with PR status after adjusting for the prognostic factors found in version 2.2. Separate models were derived for oestrogen receptor (ER)-negative cases and ER-positive cases. Data from an independent cohort of 11,365 breast cancer patients from New Zealand were utilised for external validation. Model calibration, discrimination and reclassification were used to test the model performance. Data from 34,265 ER-positive cases and 5,484 ER-negative cases diagnosed from 2000 to 2017 in the regions served by the Eastern cancer registry were used for model development of the updated version of PREDICT Breast. Various statistical fitting methods were applied to enhance the ability to capture the shape of the survival data and examine for any non-linear effects in the continuous prognostic factors, and to improve model performance relative to the previous version (v 2.2). These techniques were compared with each other based on the Akaike Information Criterion (AIC) value. Subsequently, Cox proportional hazards models with the optimal modelling method were fitted to estimate the prognostic effects of the risk factors found in PREDICT Breast and to compute the baseline hazard functions for the ER-specific cases separately. For external validation, data from West Midlands cancer registry on 32,408 breast cancer patients were used to determine the discriminative power, calibration and reclassification of the new version of PREDICT Breast. Findings: Having a PR-positive tumour was associated with a 23% and 28% lower risk of dying from breast cancer for women with ER-negative and ER-positive breast cancer, respectively. The area under the ROC curve increased with the addition of PR status from 0.807 to 0.809 for patients with ER-negative tumours (p = 0.023) and from 0.898 to 0.902 for patients with ER-positive tumours (p = 2.3×10−6) in the New Zealand cohort. Model calibration was modest with 940 observed deaths compared to 1,151 predicted. The AIC measurements showed that multiple fractional polynomials best capture the non-linear effects of the continuous risk factors and were used to estimate the non-linear transformations. The new model shows to be well-calibrated. 10-year breast cancer deaths were slightly under-predicted in the Eastern cancer registry (ER-: -2.5%, ER+: -1.4%) and over-predicted in the West Midlands data (ER: 0.1%, ER+: 6.8%). The AUC for 15-year breast cancer survival improved from 0.833 to 0.836 (p = 2.3x10−4) in the Eastern cancer registry data and from 0.810 to 0.812 (p = 0.098) in the West Midlands data. Conclusion: Incorporating the prognostic effects of PR status and year of diagnosis, updating the prognostic effects of all risk factors and amending the baseline hazard functions have led to an improvement of model performance of PREDICT Breast and resulted in more accurate absolute treatment benefit predictions for individual patients. I developed an enhanced version of the prognostic tool and successfully validated it utilising several independent data sets. The updated version will shortly be implemented online.