In this paper, we present a symbol-level iterative sourcechannel decoding (ISCD) algorithm for reliable transmission of variable-length codes (VLCs). Firstly, an improved source a posteriori probability (APP) decoding approach is proposed for packetized variable-length encoded Markov sources. Also proposed is a recursive implementation based on a three-dimensional joint trellis for symbol decoding of binary convolutional codes. APP channel decoding on this joint trellis is realized by modification of the Bahl-Cocke-Jelinek-Raviv algorithm and adaptation to the nonstationary VLC trellis. Simulation results indicate that the proposed ISCD scheme allows to exchange between its constituent decoders the symbol-level extrinsic information and achieves high robustness against channel noises.