University of Twente Student Theses
Estimating the risk of locoregional recurrence and second primary breast cancer: the role of distant metastasis as a competing risk
Boone, A.F. (2023) Estimating the risk of locoregional recurrence and second primary breast cancer: the role of distant metastasis as a competing risk.
Full text not available from this repository.
Full Text Status: | Access to this publication is restricted |
Embargo date: | 1 January 2025 |
Abstract: | Purpose Follow-up after breast cancer treatment focuses on early detection of locoregional recurrences (LRR) or second primary breast cancer (SP) to improve patient outcomes. Estimating the patients individual 5-year recurrence risk can help healthcare providers and patients develop personalised risk-based follow-up pathways. As the diagnosis of distant metastasis (DM) likely affects the risk of detecting LRR or SP, it should be considered as a competing risk when developing models to predict LRR or SP. The objective of this study is to assess the role of DM as a competing risk when predicting the 5-year recurrence risk for LRR and SP. Methods Data from 13,494 breast cancer patients were used. Two models were created for both outcomes LRR and SP, one where DM is considered a competing risk and one where DM is not considered a competing risk. The statistical approaches Cox regression analysis (COX) and random survival forest (RSF) were used to develop the prediction models. The predictive performance was assessed on model calibration and discrimination, absolute, mean, and relative predicted risks to assess the impact of including DM as a competing risk. Results The RSF approach showed a 5-year AUC of 0.76 for predicting LRR and DM was considered a competing risk and also when DM was not considered a competing risk. The 5-year AUC for predicting SP is 0.71 when DM was considered a competing risk and 0.70 when DM was not considered a competing risk. The mean risk difference for the predicted 5-year risk with the RSF approach was 0.07% and 0.04% for models predicting LRR and SP, respectively. For the COX approach, this was 0.11% and 0.04% for models predicting LRR and SP, respectively. In both groups, the mean predicted risk for the unknown DM cohort was lower than the mean predicted risk for the known DM cohort. The models calibration and discrimination appear largely uniform in both situations. Conclusions The mean predicted risks for all patients and model performances are almost similar in both models, DM has no substantial role in predicting the risk for LRR and SP. |
Item Type: | Essay (Master) |
Faculty: | TNW: Science and Technology |
Subject: | 44 medicine |
Programme: | Health Sciences MSc (66851) |
Link to this item: | https://purl.utwente.nl/essays/97279 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page