Malignant pleural effusion (MPE) is a useful specimen allowing for the evaluation of EGFR status in nonsmall cell lung cancer (NSCLC). However, direct sequencing of genomic DNA from MPE samples was found not to be sensitive for EGFR mutation detection. To test whether EGFR analysis from RNA is less prone to interference from nontumour cells that have no or lower EGFR expression, we compared three methods (sequencing from cell-derived RNA versus sequencing and mass-spectrometric analysis from genomic DNA), in parallel, for EGFR mutation detection from MPE samples in 150 lung adenocarcinoma patients receiving first-line tyrosine kinase inhibitors (TKIs). Among these MPE samples, EGFR mutations were much more frequently identified by sequencing using RNA than by sequencing and mass-spectrometric analysis from genomic DNA (for all mutations, 67.3 versus 44.7 and 46.7%; for L858R or exon 19 deletions, 61.3 versus 41.3 and 46.7%, respectively). The better mutation detection yield of sequencing from RNA was coupled with the superior prediction of clinical efficacy of first-line TKIs. In patients with acquired resistance, EGFR sequencing from RNA provided satisfactory detection of T790M (54.2%). These results demonstrated that EGFR sequencing using RNA as template greatly improves sensitivity for EGFR mutation detection from samples of MPE, highlighting RNA as the favourable source for analysing EGFR mutations from heterogeneous MPE specimens in NSCLC.