OBJECTIVES: Immortal time bias (ITB), the inclusion of person-time during which the study outcome cannot occur, has been shown to bias study findings.  We examine the impact of ITB by estimating the effect of chemotherapy on overall survival, and demonstrate how landmark analysis can correct the bias. METHODS: Retrospective study using the MarketScan® Research Databases with commercially and Medicare insured individuals linked to the Social Security Administration Death records. Subjects with newly diagnosed metastatic breast cancer (ICD-9-CM 174.x plus additional codes 196.xx-199.xx) and ≥1 year of continuous enrollment prior to breast cancer diagnosis were identified. Chemotherapy exposure was defined as ≥3 chemotherapy claims following metastatic cancer diagnosis. Landmark analysis was used to estimate survival rates conditional on surviving to certain time points to adjust for ITB.  Time to death or censoring was determined for the full sample and patients who survived 1, 3, 6 and 12 months. RESULTS: A total of 5759 metastatic breast cancer patients were identified of which 2932 had ≥3 claims for chemotherapy during follow-up. Average survival time for chemotherapy patients was 9.0 months longer than patients with <3 chemotherapy claims. The difference in survival times between patients with and without chemotherapy decreased as patients were required to survive for longer periods of time: 1-month survival = +8.9 months, 3-month survival = +7.0 months, 6-month survival = +6.7 months, 12-month survival = +7.0 months. The artificially increased effect of chemotherapy in the full sample analysis was due to the time between metastatic cancer diagnosis and third chemotherapy claim being “immortal” for the chemotherapy patients (median 2.6 months). CONCLUSIONS: Landmark analysis can be used to account for immortal time bias in oncologystudies analyzing the effect of new treatments or the comparative effectiveness of current treatments. However, an appropriate landmark must be chosen as results can be affected.