A combination of routine laboratory findings and vital signs can predict survival of advanced cancer patients without physician evaluation: a fractional polynomial model
Jun Hamano,Ayano Takeuchi,Takuhiro Yamaguchi,Mika Baba,Kengo Imai,Masayuki Ikenaga,Yoshihisa Matsumoto,Ryuichi Sekine,Takashi Yamaguchi,Takeshi Hirohashi,Tsukasa Tajima,Ryohei Tatara,Hiroaki Watanabe,Hiroyuki Otani,Hiroka Nagaoka,Masanori Mori,Yo Tei,Shuji Hiramoto,Tatsuya Morita
DOI: https://doi.org/10.1016/j.ejca.2018.09.037
Abstract:Introduction: There have been no reports about predicting survival of patients with advanced cancer constructed entirely with objective variables. We aimed to develop a prognostic model based on laboratory findings and vital signs using a fractional polynomial (FP) model. Methods: A multicentre prospective cohort study was conducted at 58 specialist palliative care services in Japan from September 2012 to April 2014. Eligible patients were older than 20 years and had advanced cancer. We developed models for predicting 7-day, 14-day, 30-day, 56-day and 90-day survival by using the FP modelling method. Results: Data from 1039 patients were analysed to develop each prognostic model (Objective Prognostic Index for advanced cancer [OPI-AC]). All models included the heart rate, urea and albumin, while some models included the respiratory rate, creatinine, C-reactive protein, lymphocyte count, neutrophil count, total bilirubin, lactate dehydrogenase and platelet/lymphocyte ratio. The area under the curve was 0.77, 0.81, 0.90, 0.90 and 0.92 for the 7-day, 14-day, 30-day, 56-day and 90-day model, respectively. The accuracy of the OPI-AC predicting 30-day, 56-day and 90-day survival was significantly higher than that of the Palliative Prognostic Score or the Prognosis in Palliative Care Study model, which are based on a combination of symptoms and physician estimation. Conclusion: We developed highly accurate prognostic indexes for predicting the survival of patients with advanced cancer from objective variables alone, which may be useful for end-of-life management. The FP modelling method could be promising for developing other prognostic models in future research.