Goal: The factors underlying the choice of opioids for cancer patients in primary care are largely unknown. Our aim was to describe cancer patients' first treatment episode with opioids in relation to disease characteristics and clinical course. Patients and methods: During 1997 and 1998, a population-based cohort of 4,006 incident cancer patients from a Danish county was identified. The patients were followed up from diagnosis to death or until 31 December 2003, and data on their use of opioids were obtained from a prescription database. Main results: Eventually, 54% of the cancer patients became incident users of opioids. Opioid treatment was initiated close to the diagnosis date in 20% of the patients. Most incident users (57%) were not terminal when they began using opioids, and 44% survived the first treatment episode. Of those who died, 70% received opioids in their terminal phase. The incidence rates of new opioid users were inversely related to the 5-year cancer survival period. A weak opioid was the first choice in 64% of the non-terminal users and in 43% of the terminal ones. No statistically significant differences in opioid use were found between men and women. Conclusions: Opioid use in cancer patients was not confined to the terminal course. Treatment with opioids should be viewed as a dynamic condition, with patients shifting between periods of use and non-use. The aggressiveness of the cancer and the presence of metastases were characteristics found to be strong determinants of opioid use.