We consider a semiparametric mean residual life model for regression analysis of patient’s remaining life with a long-term survivor or cure fraction in therapeutic clinical trials. An EM algorithm is developed for estimation. Theoretical asymptotic results, including identifiability of the model, are derived and small sample properties are verified via simulation. We apply the proposed model to an adjuvant therapy trial in advanced melanoma and a trial of first-line therapy for acute myeloid leukemia. The model provides an alternative interpretation of the trial results beyond the usual protocol-specified log-rank tests, or the Cox regression models.