Soft tissue sarcomas (STS) are rare cancers with a high risk of recurrence, therefore requiring structured and effective follow-up strategies. Current follow-up guidelines, such as those provided by the NCCN and ESMO, use fixed time intervals and do not fully account for individual patient risk factors when scheduling follow-up visits. Additionally, adherence to these guidelines is inconsistent, with some patients receiving unnecessary follow-up visits while others do not receive sufficient monitoring. This inefficiency places a burden on healthcare resources while potentially impacting patient outcomes.
This research aimed to develop a decision-support tool, in the form of a Bayesian network, that helps to personalise follow-up strategies. It does so by combining insights from historical patient data and expert knowledge to estimate recurrence risk, where risk is based on both likelihood of a recurrence and the severity of a potential recurrence and its treatment. The tool shows how frequent similar patients were historically followed-up and what their outcomes were. Rather than giving recommendations, the model supports clinicians by providing insights into recurrence patterns and follow-up intensity among similar cases, helping clinicians reflect on what might be appropriate for a given case.
The study consisted of three phases: identifying key recurrence risk factors and follow-up patterns through data analysis and literature review, developing and validating the Bayesian model in consultation with clinical experts and finally analysing mortality and economic effects for different follow-up frequencies.
Tumour grade and surgical margin were found to be most important risk factors for recurrence, with grade being the strongest. Two main model structures were made, one including both tumour grade and surgical margin to estimate probability of recurrence, while the other included only tumour grade as a reflection of probability of recurrence. Both models contained the variables expected remaining lifespan and fitness of the patient to estimate consequence of recurrence. The models showed considerable variation in risk classification compared to expert opinion and tumour grade alone. Since there is not one ‘correct’ way to classify risk, this reflects the perceptive nature of risk and the different models offer different viewpoints. Additionally, the study found no clear survival benefit from more intensive follow-up in any of the models, while it did substantially increase costs, especially per survivor. This highlights the need for clinicians to reflect on and discuss optimal allocation of follow-up resources.
By developing a structured, evidence-based approach to follow-up, this study contributes to improving STS care. The proposed decision-support tool could help clinicians make more transparent, risk-based follow-up decisions, while reducing unnecessary burdens on healthcare systems.