This paper summarises evidence for medicinal uses of opioids; harms related to the extra-medical use and dependence upon these drugs, and for a wide range of interventions to address the harms related to extra-medical opioid use. Finally, we use mathematical modelling to estimate harms and explore the overall health benefits of opioid agonist treatment (OAT) in a range of settings that vary in levels of opioid use and associated harms (overdose, HIV, HCV, suicide, accidental injuries) and responses. Estimates in 2017 suggest 40.5 million people were dependent upon opioids (40.5 million people, 95%UI 34.3–47.9 million) and 109,500 people died from opioid overdose (10.5,800–113,600). OAT can be highly effective in reducing illicit opioid use and improving multiple health and social outcomes, including reduced overall mortality and key causes of death including overdose, suicide, and other injuries. Modelling suggested scaling-up and retaining people in OAT, including providing OAT in prison, could avert a median of 7.7%, 14.5% and 25.9% deaths over the next 20 years (compared to scenarios without OAT) in Kentucky, Kyiv and Tehran, with more impact achieved in Tehran and Kyiv due to the added benefits on HIV mortality.. Other pharmacological and non-pharmacological treatments have varying levels of evidence for effectiveness and patient acceptability. Other effective interventions are those focused on preventing harms associated with problematic opioid use. Despite strong evidence for the effectiveness of a range of interventions to improve the health and well-being of people who are dependent on opioids, coverage is low even in high income countries. Treatment quality may be less than desirable, and considerable human, social, and economic harms arise from the criminalisation of illicit opioid use and dependence. Alternative policy frameworks are recommended that adopt a human rights and public health-based approach, do not make drug use a criminal behaviour and seek to reduce drug related harm at the population level.