Background: On average, attrition in trials of anti-obesity drugs is one third. There are many ways to analyse data sets with missing data, which leaves room for a variety of effect estimates. Intention-to-treat analysis needs imputation and a common method is to use the last observation carried forward (LOCF), although this can be problematic. Additional analyses are usually available in the unpublished clinical study reports that have been submitted to the regulatory authorities in the registration application. Objectives: We aimed to explore the robustness of the results in trials on weight loss by comparing results from unpublished reports submitted to the Danish Medicines Agency (DMA) and their corresponding published reports. Methods: We included trials on obese patients randomised to placebo or sibutramine in approved doses. We searched for published reports electronically and by inquiries to the producer. Results: We included 21 DMA reports and in 9 (43%) cases we identified a publication. The median trial duration was 12 weeks. At abstract submission, we had only extracted data on 4 (19%) of the DMA reports and the corresponding published versions. The median number of datasets analysed was 4.5 and 1 in the DMA and published reports, respectively. The DMA reports contained analyses of completers (n = 4) and available data (n = 3) and used LOCF (n = 4) and more advanced methods for imputation (n = 2), but the published reports only reported LOCF analyses (n = 4) and completers’ analysis (n = 1). In general, the LOCF analyses were conservative, but in one trial the LOCF analyses found the greatest treatment effect of sibutramine, and only these were reported in the publication. Conclusion: LOCF is not always conservative and an analysis with a reliable imputation method should be published. Final results and conclusion will be presented at the colloquium.