Three errors occurred in the paper by Wim van den Brink and colleagues (9 August, pp 310-2). In the last sentence of the results section the authors' original word “none” inexplicably lost its first “n” in the editing process and ended up as “one.” The sentence should read: “There were three deaths (one in group A, one in group B, and one in group C (in the first phase before heroin was prescribed)), none of which were related to the coprescribed heroin.” In the last sentence of the first paragraph of the discussion in the full version of the paper (on bmj.com), the proportion of participants who did not respond to the coprescribed heroin was 45-78% [not 45-88%]. Finally, in the abridged version, an electronic glitch led to the name and position of the last author (Jan M van Ree, professor) “dropping off.”