Friday, December 12, 2008

Cancer survival prediction model

Journal of Clincial Oncology has a paper discussing the development of a survival-prediction model for cancer patients. The paper actually looks at the development and validation of two models (based on the same data sets). The data mostly come from Canadian patients referred to a radiotherapy center for palliative radiotherapy (this is a large, prospectively gathered data set which includes cancer type, performance status, types of metastases, etc.). One of the models used a point system (based on type of cancer, performance status, and presence of non-bone metastases), the other model was a simple 1, 2, 3 stratification of patients based on if they had 1, 2, or 3 poor prognostic risk factors: cancer other than breast, Karnofsky score less than or equal to 60 or ECOG greater than or equal to 2, and non-bone metastases.

Those who had all 3 risk factors had a median survial of 9 weeks, with 3, 6, and 12 month survival of 35-45%, 14-27%, and 3-11% respectively (the range there is due to differences between the derivation and validation sets). Those with 2 risk factors had a median survival ~25 weeks with the 3/6/12 spread of 68-76%/45-52%/23-26%. There's a lot of extra data here - comparing the two scoring systems with each other and a previous one, etc: they all ended up having similar prognostic power.

They don't give enough information in the paper to determine how this would match up to simple time prediction based on one factor (ie performance status), nor is one ever certain how to generalize these sorts of data, particularly as this is based on patient characteristics at the time of referral for palliative radiotherapy. That said, in total over 1000 patients were included in the analyses in this paper, which is a lot for this type of research, and it probably is as good of a representative sample as anything out there. What is particularly helpful clinically, at least for me, is that it presents both median survival data, as well as, e.g., 6 month survival - getting a sense of that spread is always welcomed.

ResearchBlogging.orgE. Chow, M. Abdolell, T. Panzarella, K. Harris, A. Bezjak, P. Warde, I. Tannock (2008). Predictive Model for Survival in Patients With Advanced Cancer Journal of Clinical Oncology DOI: 10.1200/JCO.2008.17.1363

Pallimed | Blogger Template adapted from Mash2 by Bloggermint