A Bayesian meta-analysis of the prognostic value of circulating HER2/neu levels in breast cancer (BC) patients.

Abstract

643 Background: Many reports have shown the prognostic value of HER-2 measured in tumor tissue or in blood. In 2009, Finn et.al. showed that in a study of 579 MBC patients whose serum HER-2 levels (sHER-2) were constantly below normal had a longer PFS than patients with a sHER-2 level constantly above normal. Patients who converted from above normal to less than normal during therapy had a longer PFS than the opposite change. Methods: We performed a Bayesian meta-analysis of 12 studies where all sHER-2 levels were measured using an FDA cleared sHER-2 test with a standard cutoff of 15ng/ml. We chose a Bayesian approach because a “meta-analysis” is a natural extension of the Bayesian view that current knowledge is the result of prior knowledge modified by the data. After an in depth literature search, we selected 12 publications based on the following criteria. Baseline levels were available from either early stage or late stage patients who had at least a 2 year disease free or progression free survival as indicated by a Kaplan-Meier (K-M) curve. Results: The analysis included 4030 BC patients of which 1106 patients had baseline levels < 15ng/ml and 2924 patients had baseline values >15ng/ml. From the K-M curves, we estimated the number of recurrences up to 24 months in each group and prepared a 2x2 table for each study. We determined the odds ratio (OR) for the 12 studies which ranged from 0.57 to 74. A posterior distribution for the aggregated 12 studies can be represented by a Dirichlet distribution. 10,000 estimates of the aggregated OR indicated that there is a 95% credibility that the odds of a woman with baseline sHER-2 >15ng/ml recurring at or before two years is between 3.39 and 4.57 times higher than the odds of a woman whose baseline sHER-2 was < 15ng/ml recurring at or before two years. Conclusions: This meta-analysis agrees with previous studies that serum HER-2 levels > 15ng/ml can be a strong prognostic indicator for women with Breast Cancer and that managing therapy regimens to maximize the decrease in serum HER-2 levels could be important target in treating patients.