The association between long-term variation of postdiagnosis platelets and survival of pancreatic cancer (PC) has never been discussed by using dynamic survival analysis method. In this retrospective study, we analyzed 311 histologically confirmed PC patients identified from a mega population-based electronic inpatients database from 2012 to 2013 in China. Counting process approach was applied to restructure the original survival data, the association between post-diagnosis platelet count and overall survival (OS) of PC was evaluated by multiple failure-time Cox proportional hazards model. After counting process adjustment, multiple failure-time Cox proportional hazards model revealed that, regardless of the treatment modalities PC patients received, postdiagnosis thrombocytopenia was prominently associated with OS, compared with PC patients with normally ranged platelet count, the HRs ranged from 2.04 (95% CI: 1.14-3.67) to 10.82 (95% CI: 2.63-44.54), and this inverse association was robust based on further sensitivity analysis. On the contrary, the association between thrombocytosis and OS of PC tended to be inconclusive. Our findings suggested that postdiagnosis thrombocytopenia was associated with significantly compromised survival among PC patients from this large retrospective cohort. Underlying mechanisms behind this association should be further investigated.
Keywords: Counting process approach; pancreatic cancer; platelet count; survival.
© 2016 The Authors. Cancer Medicine published by John Wiley & Sons Ltd.