Advantages of a multi-state approach in surgical research: how intermediate events and risk factor profile affect the prognosis of a patient with locally advanced rectal cancer

Abstract Background Standard survival analysis fails to give insight into what happens to a patient after a first outcome event (like first relapse of a disease). Multi-state models are a useful tool for analyzing survival data when different treatments and results (intermediate events) can occur. Aim of this study was to implement a multi-state model on data of patients with rectal cancer to illustrate the advantages of multi-state analysis in comparison to standard survival analysis. Methods We re-analyzed data from the RCT FOGT-2 study by using a multi-state model. Based on the results we defined a high and low risk reference patient. Using dynamic prediction, we estimated how the survival probability changes as more information about the clinical history of the patient becomes available. Results A patient with stage UICC IIIc (vs UICC II) has a higher risk to develop distant metastasis (DM) or both DM and local recurrence (LR) if he/she discontinues chemotherapy within 6 months or between 6 and 12 months, as well as after the completion of 12 months CTx with HR 3.55 (p = 0.026), 5.33 (p = 0.001) and 3.37 (p < 0.001), respectively. He/she also has a higher risk to die after the development of DM (HR 1.72, p = 0.023). Anterior resection vs. abdominoperineal amputation means 63% risk reduction to develop DM or both DM and LR (HR 0.37, p = 0.003) after discontinuation of chemotherapy between 6 and 12 months. After development of LR, a woman has a 4.62 times higher risk to die (p = 0.006). A high risk reference patient has an estimated 43% 5-year survival probability at start of CTx, whereas for a low risk patient this is 79%. After the development of DM 1 year later, the high risk patient has an estimated 5-year survival probability of 11% and the low risk patient one of 21%. Conclusions Multi-state models help to gain additional insight into the complex events after start of treatment. Dynamic prediction shows how survival probabilities change by progression of the clinical history