OBJECTIVES: In light of lacking evidence, persistence is an important component of assessing opioid safety and effectiveness. This study aims to evaluate different patterns of persistent opioid use in patients with cancer and non-cancer pain.  METHODS: This cross sectional study was conducted at an outpatient pharmacy Hospital Tengku Ampuan Afzan Kuantan (HTAA) Malaysia. Prescription records for four opioids (dihydrocodeine, fentanyl, morphine and oxycodone) issued between January 2013 and December 2015 were examined.  Persistence pattern is defined based on different levels of dispensed opioid amounts, number of prescriptions and the number of quarters out of the year in which prescriptions were dispensed. The strict definition describes a typical patient using opioids to achieve a continuous serum concentration in the therapeutic range, the intermediate definition represents a typical patient using opioid daily but not around the clock and the wide definition describes a typical patient who uses opioids for at least half of the days in a year. Patients who did not meet any of the definitions were categorized as non-persistent. Number of patients for each persistence group was calculated and all analyses were performed using Stata 13.  RESULTS: A total of 1,684 opioid prescriptions for 379 patients were dispensed during the study period. Seventy five percent of (n=283/379) patients were associated with cancer diagnosis while 25% (n=96/379) with non-cancer diagnosis. The prevalence of persistent opioid users was higher in the non-cancer group (13.5%) compared to the cancer group (9.9%). The prevalence of strict, intermediate, and wide definitions were (23.1% vs. 28.6%), (15.4% vs 10.7%), and (61.5% vs 60.7%) in the non-cancer and cancer groups respectively. CONCLUSIONS: Persistent opioid users were more prevalent in the non-cancer group with the majority in the wide and intermediate persistence patterns. Further investigation is required to examine the clinical outcomes related to differential patterns of persistence.