Back

Research designs and statistical methods in medical abstracts

Abstract number: p1511

Kompoti  M., Matsagoura  M., Koutsovasilis  A., Koutsovasili  A., Drimis  S.

Statistical methods used in biomedical research articles are being increasingly scrutinized in medical journals. However, no such strict policy is generally applied in abstracts presented in medical congresses.

Objective: 

This study aimed at assessing the frequency of research designs and statistical methods reported in abstracts presented in two successive years of the European Congress of Clinical Microbiology and Infectious Diseases (ECCMID).

Material and methods: 

We reviewed all abstracts included in the abstract book of the 14th ECCMID (Prague 2004) (PG) and the 15th ECCMID (Copenhagen 2005) (CP). All abstracts of original research studies but no abstracts of lectures were included in our study. Two independent investigators read all abstracts and extracted information concerning origin, type (clinical, laboratory, animal model), research design, sample size and statistical methods used in the study. Data analysis was performed with logistic regression and Pearson's chi-square test for categorical variables and Student's t-test for continuous variables. Statistical significance level was set at p < 0.05.

Results: 

A total of 4178 abstracts were included in the analysis according to eligibility criteria (2110 from PG and 2068 from CP). Laboratory studies prevailed (61%) followed by clinical studies (35%) and experimental studies with animal models (4%). The majority (79.3%) of the studies were observational (retrospective, prospective, cross-sectional) of which 6.3% concerned diagnostic accuracy testing of laboratory methods and 2.3% were pharmacological studies, 3.8% were randomized controlled trials. Statistical evaluation was clearly described in 26.5% of abstracts (26.1% in PG and 19.8% in CP, p < 0.001), while the rest of abstracts included only descriptive statistics or no statistics at all. The proportion of statistical methods reporting varied according to the type of the study (animal model studies 49.4%, clinical studies 38.9% and laboratory studies 12.0%, p < 0.001). Multicentre research studies reported statistics more frequently than single-center studies (26.5% vs. 21.5%, respectively, p = 0.005).

Conclusions: 

Statistical analysis is an inseparable part of original research. Research design as well as the implemented statistical methods should always be reported in an adequate manner, thus improving the scientific quality of abstracts.

Session Details

Date: 01/08/2007
Time: 00:00-00:00
Session name: XXIst ISTH Congress
Subject:
Location: Oxford, UK
Presentation type:
Back to top