Forest biomass retrieval using P-band synthetic aperture radar (SAR) data has been studied during the past two decades. However, much progress remains necessary to recover biomass for tropical dense forests, where above-ground biomass (AGB) values are often found between 250 and 400 t/ha. In this high range, the sensitivity of radar backscatter to AGB is small and can be obscured by disturbing effects. These are related to measurement errors, forest characteristics, and environment variations, but the most important is the underlying topography. In this work, we develop a new backscatter coefficient called $t^0$ in order to minimize topography effects, based on 1) the medium nonreflection symmetry related to azimuth slopes and 2) the change in backscatter angular variation coming with the change in scattering mechanisms due to slopes. The method is tested on TropiSAR campaign data acquired over tropical forest plots in French Guiana. Using power laws between AGB and the backscattering coefficients, results based on the correlation analysis demonstrate the relevance of the proposed method.