In the context of personalized oncology, screening for somatic tumor mutations is crucial for prediction of an individual patient’s response to therapy. Massive parallel sequencing (MPS) has been suggested for routine diagnostics, but this technology has not been sufficiently evaluated with respect to feasibility, reliability, and cost effectiveness with routine diagnostic formalin-fixed, paraffin-embedded material. We performed ultradeep targeted semiconductor-based MPS (190 amplicons covering hotspot mutations in 46 genes) in a variety of formalin-fixed, paraffin-embedded diagnostic samples of lung adenocarcinoma tissue with known EGFR mutations (n = 28). The samples reflected the typical spectrum of tissue material for diagnostics, including small biopsies and samples with low tumor-cell content. Using MPS, we successfully sequenced all samples, with a mean read depth of 2947 reads per amplicon. High-quality sequence reads were obtained from samples containing ≥10% tumor material. In all but one sample, variant calling identified the same EGFR mutations as were detected by conventional Sanger sequencing. Moreover, we identified 43 additional mutations in 17 genes and detected amplifications in the EGFR and ERBB2 genes. MPS performance was reliable and independent of the type of material, as well as of the fixation and extraction methods, but was influenced by tumor-cell content and the degree of DNA degradation. Using sample multiplexing, focused MPS approached diagnostically acceptable cost rates.