All 154 oral abstracts from the 1st World Conference on Sports Injury Prevention in Oslo 2005 were analyzed in a database. PubMed and Medline online searches were performed as continuous follow up over a 42-month period to ascertain whether the presented oral abstracts were followed by full paper publication. These searches were done using the authors names (first, second, and last authors, each searched individually); then they were compared with probable keywords and title phrases of the abstract. When full paper publications were identified, the title of the full paper publication, the name(s) of authors, the location(s) where research was conducted, study design, sample size, follow-up duration, number of withdrawals, and study results data were compared with the information available from the conference abstracts. Verification of conference abstracts and their corresponding full paper was performed. The following items were verified: same title and author name(s); same location(s) for research; consistent study designs, follow-up durations, and numbers of withdrawals; and no major differences in study results data. Multiple full paper publications for a single abstract were also identified; just the first full paper publication was used for analysis.
For each conference abstract and its corresponding full paper publication, CONSORT abstract criteria for all randomized (Additional file 1: Appendix 1), controlled trials and STROBE for all observational studies (Additional file 1: Appendix 2) were applied. Scores were calculated for 17 CONSORT criteria and 22 STROBE criteria for all conference abstracts; two investigators independently assessed each conference abstract and the corresponding full paper publication abstract and calculated the score. The mean score of both investigators in each abstract was taken. Inter-rater variability was 6%.
Clinical studies with known study designs were separated from experimental studies. A randomized controlled trial was identified if the abstract described a prospective study in which individuals were allocated at random to an intervention or a control group. Observational studies included cohort, case-control, and cross-sectional studies; systematic reviews; and case series and case reports. If the study design was still unclear after attempts to match it with the formats outlined above, abstracts were analyzed by both researchers together (UY and KK). Among the facets of each study that were re-analyzed were how the study was performed; patient recruitment; study duration and setting; intervention and follow up; statistical data, analysis and data in the results. Furthermore, the entire abstract was searched for specific words like "prevalence", "placebo", "blinded", "random", "questionnaire", "lab techniques (PCR, Western/Southern/Northern-blot)", "odds ratio", and "relative risk" to determine the study design.
In addition to the above, the names and numbers of authors; the numbers of centers (location where research was conducted); the objective and/or hypothesis; the author's interpretation of data; sample size and study results; standard deviations (SD) and confidence intervals (CI) (actual numbers); location(s) of study; and method of study (randomized controlled trial, observational trial, statistical analysis, primary outcomes) were all recorded. For randomized controlled trials, follow-up duration, method of randomization, blinding, use intent-to-treat analysis, and number of withdrawals/dropouts were also noted.
The scores of 17 CONSORT criteria and 22 STROBE criteria for reporting were determined. Publication rate, publication time, minor and major inconsistencies between conference abstracts, and whether the study was ultimately published as a full paper were measured. Minor inconsistencies included differences in title, authors, research center, presentation of all outcomes (p-value, confidence interval, Pearson), and authors' interpretation of data (conclusion). Major inconsistencies included discrepancies in study objective and/or hypothesis, study design, primary outcome measures, sample size, statistical analysis, results, and standard deviations/confidence intervals. The primary outcome was defined as the main outcome reported in an abstract. If the amount of reported result data in the abstract did not match the full paper it was described as "results different". Mismatches in the amount of study result data presented between the abstract and the full paper were reported in four categories. If the full paper reported fewer result data than were presented in the abstract, it was reported as "data deleted". A full paper with more result data than were present in the abstract was reported as "data added"; if completely result different data were reported in the full paper in comparison to the abstract, it was described as "results completely changed". For the specific instance in which just the standard deviation, confidence interval, injury rate, or incidence changed, it was described as "SD/CI, Injury rate, Incidence changed". The impact factor, which is dependent on the year and the journal's distribution of the published abstracts, was also determined.
Descriptive statistics consisted of the calculation of frequencies and percentages. Clopper and Pearson analysis were used for binomial proportion confidence interval. The T-test was used to compare minor and major inconsistencies, the CONSORT and STROBE score, and the impact factors between randomized, controlled trials and observational studies. Chi-squared tests were used to compare publication rates, odds ratios, 95% confidence intervals, and p values according to the different countries. Statistical significance was defined as p < 0.05. Data were analyzed using the SPSS statistical software package Version 14.0 and StatXact version 6.