From conference abstract to full paper: differences between data presented in conferences and journals

Abstract number: 1134_02_74

Rosmarakis E., Vergidis P., Kapaskelis A., Paraschakis K., Falagas M.

Background:

On some occasions, we noted differences between data presented in conference abstracts and subsequent published papers. We studied the frequency and the type of these differences in the fields of Infectious Diseases and Microbiology.

Methods:

We reviewed all abstracts from the first session of 7 out of 15 major research categories presented in the 1999 and 2000 ICAAC. For each selected pair of ICAAC abstract and related full paper published in journals indexed by Index Medicus, two independent investigators performed a comparison of all data in the abstract and the corresponding information in the published paper. Using Cox logistic regression models we analysed variables for association with differences of data.

Results:

From 190 abstracts that were reviewed, 68 (35.8%) were subsequently published as papers by March 2004. From them, 52 referred to the same study period and population for both the abstract and the paper. Differences between data presented in conference abstracts and published papers were found in 28 out of 51 pairs which were analysed further (54.9%, 95% C.I.: 41.2%–68.6%). The identified differences were related to the numbers and/or rates of the studied patients (11/28), numbers or rates of isolates (9/28), MICs values or Ki values (5/28), other chemical properties of antibiotics (2/28), odds ratio (1/28), and duration of observation (1/28). The differences were substantial in several pairs. Time to publication of full paper was found to be independently associated with presence of differences (p = 0.029, OR = 2.043 per year), while the research category, type of presentation (oral or poster), number of publications of the presenting and last author, impact factor of the journal, and country of origin were not.

Conclusions:

While there are several explanations for the noted differences between data presented in conference abstracts and full papers, it is likely that the research community may improve the accuracy of presentation of data.