Background and purpose: The European Society for Therapeutic Radiology and Oncology was funded by the EU for a project on Recording providing Education, and Ameliorating the Consequences of Treatment (REACT). An important aim of follow-up (FU) after treatment for cancer is to detect various events associated with disease recurrence or metastatic spread or severe treatment-related complications as early as possible. Each tumour type may show a specific pattern and timing of these events related to different prognostic factors. The aim of this study was to propose a way of defining an optimal timing schedule for follow-up after treatment based on the analysis of failure patterns determined from follow-up data from prospective clinical trials.
Material and methods: Cox proportional hazards model was used to identify prognostic factors associated with each failure type (loco-regional recurrence (LR), distant metastasis (DM) or side effects (SE)). Competing risks methods were applied to estimate the cumulative incidence functions (CIF), adjusted on the significant prognostic factors. Equally spaced quantiles of the CIF were then used to estimate the corresponding optimised follow-up times depending on a pre-specified total number of visits. Follow-up data from the CHART bronchus clinical trial were used to analyse the pattern of time to first failure.
Results: A significantly higher risk of failure was observed for males (SE), stage III (DM) and conventional treatment (LR). Overall, patients treated with CHART needed 1 fewer visit in each category of patients compared to the Conventional group. For example, stage III male patients treated with CHART would need 8 visits during the first two years at 7, 11, 16, 24, 37, 52, 64 and 104 weeks rather than the 9 follow-up visits planned in the protocol. Similar patients treated with Conventional radiotherapy would need 8 visits at 3, 5, 7, 11, 15, 24, 52 and 104 weeks.
Conclusions: Use of these methods would allow timing of follow-up visits to be adapted according to tumour site and prognostic factors determined previously from audit or clinical trials. Application of this approach could optimize the timing of follow-up visits by placing them closer to the times when failures are expected to occur. It does not address the wider issues of follow-up such as who should do it or what should be done for which further studies are required.