After the introduction of abuse-deterrent OxyContin in 2010, states with widespread extramedical OxyContin use experienced steep increases in heroin deaths, implying substitution from OxyContin to heroin. Leveraging cross-state variation in initial OxyContin utilization, we show the OxyContin reformulation also induced substitution to a similar prescription opioid product, Opana ER. States with high Opana ER utilization after the OxyContin reformulation experienced continued growth in prescription opioid deaths, and after Opana ER was reformulated 18 months later, an additional wave of substitution to heroin previously solely attributed to OxyContin. Our results highlight underappreciated substitution pathways throughout these pivotal years of the epidemic.