A multistate survival model in rectal cancer surgery research for locally advanced patients

Abstract

Background: One of the most appropriate methods for analyzing longitudinal data is multistate model. This study has aimed to evaluate the risk factors of transfer to local recurrence (LR), distant metastasis (DM), and death in rectal cancer patients through multistate survival analysis. Materials and Methods: This is a retrospective cohort of rectal cancer patients in Mashhad, Iran. Multistate models were applied to show the difference between the significant risk factors affecting death and recurrence in different defined transitions. Risk factors include age, sex, primary surgical technique, tumor location, postoperative tumor stage, circumferential or distal resection involvement, surgery time, and surgical complications. Results: A total of 280 eligible patients with a median (interquartile range) survival time of 60 (42-76.2) months were investigated. Based on Cox proportional multistate model, the hazard ratio (HR) of DM increases by 3%/1-year increase in age (P = 0.018). The HR of DM and the HR of LR in patients with postoperative disease Stage II/III were 3.06 and 2.53 times higher than patients with cancer Stage 0/I (P < 0.05). When the resection margins of distal or circumferential were involved, the HR of DM was 3.58 times higher than those patients without involvement. In the extended multistate model, time of DM was a significant predictor of death (P = 0.006). Conclusion: Age and margin involvement in DM path and stage in LR and DM path had a significant effect; however, no effective variable was seen on the death of patients with recurrence. The time of metastasis also had an effect on the path of death.