Introduction Accurate and complete reporting of information about trial participants is important to apply results in clinical practice. We studied whether and how eligibility criteria (EC) of participants of randomised trials pre-specified in study protocols are reported in subsequent articles.

Methods By electronic literature searches and survey of applicants, we identified 78 full publications corresponding to 52 trial protocols submitted in 2000 to the ethics committee of the University of Freiburg/Germany. From both sources we extracted information on EC, whether they were matching, missing, modified or added in the publications.

Results We found differences between protocols and publications for all 52 trials. Information on EC was missing in publications corresponding to 52 trials (100%; 95% CI 93 to 100), modified for 44 trials (85%; 95% CI 72 to 93) or newly added for 21 trials (41%; 95% CI 27 to 55). The mean number of EC per trial was 25 (range 7–43); the mean proportion of matching EC per trial was 50% (range 13–93; 95% CI 44 to 55). Of 1248 EC pre-specified in protocols, 606 (49%) were matching, 479 (38%) missing, and 163 (13%) modified in subsequent publications. Fifty-one EC were added to publications. Most of the missing EC (96%) and about half of the modified EC (54%) suggested broader study populations, most added EC (86%) narrower study populations.

Conclusion Published trial reports generally do not reflect the definitions of study populations pre-specified in the protocol. This hampers a proper assessment of the applicability of trial results.