Objective: To assess the characteristics of medical research that is press released by general medical journals and reported in newspapers.
Design: Longitudinal study.
Data sources: All original research articles published in Lancet and BMJ during 1999 and 2000.
Main outcome measures: Inclusion of articles in Lancet or BMJ press releases, and reporting of articles in Times or Sun newspapers.
Results: Of 1193 original research articles, 517 (43%) were highlighted in a press release and 81 (7%) were reported in one or both newspapers. All articles covered in newspapers had been press released. The probability of inclusion in press releases was similar for observational studies and randomised controlled trials, but trials were less likely to be covered in the newspapers (odds ratio 0.15 (95% confidence interval 0.06 to 0.37)). Good news and bad news were equally likely to be press released, but bad news was more likely to be reported in newspapers (1.74 (1.07 to 2.83)). Studies of women's health, reproduction, and cancer were more likely to be press released and covered in newspapers. Studies from industrialised countries other than Britain were less likely to be reported in newspapers (0.51 (0.31 to 0.82)), and no studies from developing countries were covered.
Conclusions: Characteristics of articles were more strongly associated with selection for reporting in newspapers than with selection for inclusion in press releases, although each stage influenced the reporting process. Newspapers underreported randomised trials, emphasised bad news from observational studies, and ignored research from developing countries.