Pleural metastasis is one of the most common complications in lung cancers. However, no effective therapy for pleural metastasis has been established thus far. We have constructed a metastatic model of non-small cell lung cancer (NSCLC) by injecting human NSCLC cell lines directly into the left pleural cavity of BALB/c nude mice. Because this model is easy to construct and the results are reproducible, we used this model for a preclinical evaluation of gene therapy for pleural metastasis of NSCLC. We took the novel approach of in vivo lipofection of a suicidal gene to lung cancer cells metastasized to the pleural cavity. A human lung cancer cell line, PC14, was inoculated into the pleural cavity of nude mice. After 1 day, a herpes simplex virus thymidine kinase gene expression plasmid was injected intrapleurally as a DNA-liposome complex, and ganciclovir was subsequently administered for 8 days. The survival rates of the ganciclovir-treated group were significantly better than those of the control groups. Flow cytometric analysis using a green fluorescent protein expression plasmid suggested that the transfection efficiency in the pleural cavity was 13.6%. Moreover, due to a bystander effect with PC14 cells, 10% of the gene transfer efficiency was sufficient to eradicate or suppress pleural metastasis. This preclinical study suggests the therapeutic feasibility of an in vivo lipofection-based suicidal gene/prodrug strategy for pleural metastasis of NSCLC.