Meta Analysis of Advanced Cancer Survival Data Using Lognormal Parametric Fitting: A Statistical Method to Identify Effective Treatment Protocols

S.      Qazi; D.      DuMez; F.   M.   Uckun

doi:10.2174/138161207780765882

Abstract

We describe the use of a parametric lognormal model to calculate and compare survival statistics in the clinical treatment of advanced/metastatic pancreatic, breast and colon cancers. The fit using the lognormal model explained greater than 90% (R2 ranged from 0.917 to 0.998 for a total of the 51 arms from published studies) of the variation in the cumulative survival statistics of patients treated for advanced cancers. A meta-analytic Q-test was performed to test whether there were significant differences between different studies. For all three cancer types, the Q-test showed highly significant differences between the survival arms (p < 0.0001 for pancreatic, breast and colon cancers). The z-values expressed the difference of the average of lognormal means relative to each study in terms of deviation expressed in standard errors. The treatments that were most effective ranked with the highest z-value: Doxorubicin plus docetaxel for pancreatic cancer (z-value = 4.1); Capecitabine plus paclitaxel for breast cancer (z-value = 3); irinotecan, fluorouracil and folinate for colon cancer (z-value = 7.4).

Keywords: kaplan-meier (km) method, cancer-specific survival rates (cssr), lognormal model, tumor growth, bay, cetuximab, colon cancer

« Previous Next »