It has come to our attention that the design that we used in the 2 longitudinal studies on violent crime in bipolar disorder1 underestimated the relative risks of violent crime associated with patients with bipolar disorder, although the absolute risks of violent offending and substance abuse in bipolar disorder remain unchanged. Our design measured rates of violent crime in controls from age 15 years and in cases from after diagnosis. This meant that the time at risk was not similar between cases and controls. We discussed various other potential biases in our article, including the reliance on hospital data to ascertain the cases (which will likely overestimate relative risks) and conviction information to measure violence as an outcome (which will underestimate absolute risks), but would like to address this limitation in the present letter.