Sir

Like the editors of Nature (see Nature 415, 101; 2002), we too realize that it is in our best interests to carry out editorial checks of data from the Institute of Scientific Information (ISI). In 1998, The Lancet's impact factor, as calculated by ISI, dropped to 11.79 from a previous stable value of about 17 for the previous four years. In 1997, The Lancet had decided to divide letters into Correspondence (not counted in ISI's denominator) and Research Letters (peer-reviewed and containing original data, hence coded by the ISI as citable for the denominator). This division increased our number of citable items and thus we attributed most of the drop to this change, an assumption confirmed by the ISI.

Recently, we noted that the number of citable items listed for 2000 was higher (821) than informal calculation would suggest. After hand-coding each issue, we found that 684 items should form the denominator.

Meanwhile, out of editorial interest, we looked at citation data for 1999 in a file purchased from the ISI (based on papers published in 1997 and 1998). As Nature did, we also found examples of unexpectedly low numbers of citations for large trials that had only a group name in the author byline compared with those that had at least one named author. The International Stroke Trial (Lancet 349, 1569–1581; 1997), for example, was listed as having one citation in 1999. The ISI confirmed to us that group names are a potential landmine for citation accuracy and said it was in the process of developing a new program to address this issue.