Er relative risici og odds-ratioer i resumeer trovaerdige? Sekundaerpublikation
Engelsk titel: Are relative risks and odds ratios in abstracts believable? Secondary publication
Läs online
Författare:
Götzsche PC
Email: pcg@cochrane.dk
Språk: Dan
Antal referenser: 10
Dokumenttyp:
Artikel
UI-nummer: 06081673
Sammanfattning
The first result in 520 abstracts that reported relative risks or odds ratios was statistically significant in 70% of the randomized trials, 84% of cohort studies and 84% of case-control studies. Many results were derived from subgroup or secondary analyses or biased selection of results. The distribution of p values in the interval 0.04-0.06 was extremely skewed for trials, but most of the significant results were wrong, very doubtful, or could be discussed. The skewness was even larger in observational studies. This study shows that significant results reported in abstracts should be interpreted with great caution.