BMJ article challenges COPE retraction guidelines