OBJECTIVES: Ethics committees often require cross-over design for highly sensitive circumstances where it may be unethical to withhold active therapy. This can be the case particularly in oncology. Cross over can, however, dilute the treatment effect seen in trial analyses. The German Institute for Quality and Efficiency in Health Care (IQWiG) follows distinct thresholds for comparative treatment effect, which it requires for a positive early benefit assessment (EBA) rating. The upper 95% confidence interval must be at least 0.85 to receive the highest rating. We evaluated if cross-over designs may negatively affect benefit assessment in Germany.  METHODS: Oncology medicines that finished EBA procedures in Germany until June 2014 were evaluated for cross over in the manufacturer’s dossier. The extent of cross over on the observed treatment effects was investigated, as well as how the designs may affect the EBA ratings. RESULTS: Ten out of 24 EBAs in oncology included assessment of trials with cross-over design. Cross over may have affected the observed treatment effects as demonstrated by a number of examples. Firstly, the proportion of patient cross over was as high as 62% for crizotinib. For vandetanib, 12 out of 13 remissions in the control group could be attributed to a switch to active therapy. The hazard ratio for overall survival (OS) with vemurafenib versus dacarbazine was 0.37 without cross over at first data-cut in comparison to 0.62 with 24% cross over at a later timepoint. These examples suggest that cross-over designs are both present in EBAs in oncology, and may affect the extent of comparative treatment effect. The affect of cross-over design was not systematically considered by IQWiG. CONCLUSIONS: Cross-over design is an ethical necessity. However, it is known that these designs dilute treatment effect signals. German HTA EBAs need to improve in systematically accounting for such cross-over affects.