In this paper we consider the problem of maximum-likelihood decoding of a coded orthogonal frequency-division multiplexing (OFDM) transmission plagued by the presence of narrowband interference. The presence of NBI characterizes many practical contexts, including cellular applications and emerging spectrum sharing systems, where coexistence of different types of wireless services over the same frequency band may result into remarkable co-channel interference. The coding scheme is the bit-interleaved coded modulation (BICM). Conventional maximum likelihood (ML) decision strategies for BICM-OFDM transmissions suffer from significant performance degradation in the presence of NBI. To overcome this difficulty, in this work we develop a new bit-metric for the decoding of the transmitted data that achieves near-optimum results.