In this target trial emulation of 333,687 US Veterans Affairs patients with type 2 diabetes, initiation of glucagon-like peptide-1 receptor agonists was not associated with a higher 1-year risk of all-cause acute pancreatitis compared with sulfonylureas. This overall neutral result reflected opposing cause-specific effects: a higher risk of suspected drug-induced pancreatitis was offset by lower risks of alcohol-induced and hypertriglyceridemia-associated pancreatitis. These effects were time-dependent, with early clustering of suspected drug-induced events and later reductions in metabolically driven pancreatitis, resulting in a transient increase in overall risk early after treatment initiation followed by similar risk thereafter.