Characterization of published errors in high-impact oncology journals

Purpose To assess the frequency and propagation rate of published errors in the oncology literature and to determine possible contributing factors. Methods We reviewed 10 major oncology journals to determine variability in the online presentation of errata. Canadian oncologists were surveyed regarding characteristics that may influence error propagation. Errors published during 2004-2007 in the Journal of Clinical Oncology (JCO) and the Journal of the National Cancer Institute (JNCI) were classified as trivial or serious (that is, whether change in outcome was involved). The frequency of citation and error propagation was determined for serious errors. Results Of the 10 journals reviewed, 9 present links from the original article to the erratum; in 4 of those 9 journals, at least 1 link was missing. Survey results indicate that 33% of oncologists do not read errata, and 45% have read only the abstract when referencing an article. Although 59% of oncologists have noticed errors in cancer publications, only 13% reported the error. Together, Jco and JNCI published 190 errata, for an error rate of 4% +/- 1% (standard deviation) annually; 26 of 190 errors were serious (14%). The median time from publication of the article to the corresponding erratum was 3.5 months for trivial errors as compared with 8.3 months for serious errors (p = 0.03). Error propagation in citations before and after publication of the erratum was 15% and 2% respectively (p<0.01). Conclusions Error rates in high-impact oncology journals average 4%, which is likely an underestimate, because errors noticed by readers are not consistently reported. Propagation of serious errors decreases, but still continues, after publication of errata.