Sea-ice thickness at global scale is an important variable in the polar climate system. Only satellite altimeters such as onboard the CryoSat-2 mission allow us to obtain sea-ice thickness on hemispherical scale. Accurate CryoSat-2 altimeter range measurements provide surface elevations which have to be referenced to the local sea level to obtain sea-ice freeboard that can be converted into sea-ice thickness assuming hydrostatic equilibrium. The local sea-surface height is determined by careful detection of leads in the ice surface using the specific characteristics of the radar signal. Off-nadir reflections from leads can significantly affect the range retracking and hence bias the surface elevations of leads and sea ice. This can finally lead to a negative freeboard and hence also affects the thickness and volume retrieval. We present a method for the classification of CryoSat-2 radar echoes to correctly discriminate between valid and off-nadir biased echoes. We apply our classification to a CryoSat-2 track from December 15 where 50 leads over a distance of 2,300 km are identified. Overall 22 % of the surface elevations are associated with biased radar echoes.