OBJECTIVES The objective of this study was to examine the utilization patterns of cancer medications beyond their labeled indications approved by the FDA in community oncology practices. METHODS Drug prescription information from a community oncology data warehouse was used for two separate analyses. Patients were categorized according to whether they had an ICD-9 diagnosis code for one of four cancer types including lung, breast, bladder or gastric, and having no other malignancy. The frequency of use of various oncology drugs was examined for each of these groups, against a set of medications that were FDA-approved for these indications or were recommended by NCCN guidelines. In the second analysis, patients with a single malignancy, who received any of the five oncology drugs (paclitaxel, vinorelbine, rituximab, bevacizumab) and gemcitabine, were counted. Comparisons were then made against the cancer indications for which these agents were approved by the FDA. RESULTS Seventy-eight percent of breast and 95% of lung cancer patients received medications approved for these indications, while 68% and 75% also received drugs that were not approved by the FDA for those conditions. More than 99.7% of these patients received agents recommended on NCCN guidelines. None of the bladder cancer patients and only 5% of gastric cancer patients received drugs approved for these indications, while 97% and 95% of them received guideline-recommended drugs. Only half of the patients given paclitaxel or bevacizumab received these for an FDA-approved indication. In the case of vinorelbine and gemictabine, the proportion was lower at 30% and 40% respectively, while it was higher for rituximab at 60%. CONCLUSIONS Oncologists' choice of drugs is driven by evidence-based guidelines, independent of FDA approval. There is a high and varying proportion of off-label use across oncology medications and cancer types.