Combination of variables predict two-week and two-month survival for advanced cancer
FRIDAY, Aug. 26 (HealthDay News) -- A combination of clinical and laboratory variables can effectively predict two-week and two-month survival in patients with advanced cancer who are no longer being treated, according to a study published online Aug. 25 in the BMJ.
Bridget Gwilliam, from St. George's University of London in the United Kingdom, and colleagues sought to develop a novel prognostic indicator for patients with advanced cancer, which was significantly better than clinicians' estimates of survival. A total of 1,018 participants with locally advanced or metastatic cancer, who were not being treated for cancer and were recently referred to palliative care services, were included in the analysis. The model's prediction of survival for patients, in terms of days (0-13 days), weeks (14-55 days), or months (>55 days), was compared with their actual survival, and with clinicians' predictions. Separate prognostic models were used for patients with and without blood results (Prognosis in Palliative Care Study [PiPS-B and PiPS-A, respectively]).
The investigators identified 11 core variables that independently predicted both two-week and two-month survival. Four variables only had prognostic significance for two-week survival and eight had significance only for two-month survival. The area under the curve varied between 0.79 and 0.86 for all models. Actual survival and PiPS predictions had 57.3 percent absolute agreement between them. The median survival of days, weeks, and months was 5, 33, and 92 days, respectively, for PiPS-A category, and 7, 32, and 100.5 days, respectively, for PiPS-B categories. Performance of all models was either comparable to or better than clinicians' estimates of survival.
"In patients with advanced cancer no longer being treated, a combination of clinical and laboratory variables can reliably predict two-week and two-month survival," the authors write.
Full Text (subscription or payment may be required)
Editorial (subscription or payment may be required)