OBJECTIVES:: We aimed to understand trends in opioid overdoses, naloxone dosing, and overdose reversal from 2014 to 2017 at Insite, a community-supervised consumption site in Vancouver, British Columbia. METHODS:: We performed a retrospective cohort study of patients who overdosed on opioids at Insite. We evaluated yearly trends in total overdoses, naloxone doses administered, and proportions of patients reversed (based on our definitions of “probable” and “confirmed” reversal), with particular attention to trends before and after the emergence of illicit fentanyl in 2015. RESULTS:: There was an increase in total overdoses at Insite from 2014–2015 (n=586) to 2016–2017 (n=2033). Overdose reversal data were limited by a large proportion of patients for whom there was not enough information to adjudicate whether reversal occurred, and therefore were marked “unspecified” [n=1537/2619 (58.7%)]. Within these limitations, fewer patients were reversed after 2015 (44.1% in 2016, 29.7% in 2017) than before (47.8% in 2014, 55.1% in 2015) (χ=73.1, P<0.001). Despite this, naloxone doses remained unchanged between 2014–2015 and 2016–2017 {median: 0.4 mg [interquartile range (IQR): 0.4–0.8 mg] for both, P=0.21}. Insite staff administered higher doses to patients not successfully reversed [median: 0.8 mg (IQR: 0.4–0.8 mg)] compared with those reversed [median: 0.4 mg (IQR: 0.4–0.8 mg)] (P=0.021), and to patients offered transfer to hospital [median: 0.8 mg (IQR: 0.4–0.8 mg)] compared with those discharged home [median: 0.4 mg (IQR: 0.4–0.8 mg)] (Kruskal-Wallis H=288.7, P<0.001). CONCLUSION:: Overdose numbers at Insite increased and fewer patients were successfully reversed following the emergence of illicit fentanyl in 2015, but naloxone doses remained unchanged, suggesting that traditional naloxone dosing does not optimally reverse overdoses caused by fentanyl. Our study supports an alternative approach to naloxone dosing in patients with suspected ultrapotent opioid toxicity and in communities with high fentanyl prevalence.