Reference data on rats provide help. Seralini's test rats with GMO diet die at normal rates

Legend to Fig.C. Prediction intervals of level 90% and 95% survival rates at 2 years obtained from Harlan data and survival rates observed in the experimental groups and control groups. (From the French HCB report provided in an earlier post)

The back-story to this post about the disputed claims by French scientist G-E Seralini that GMO corn harms rats is given here. A new report from HCB provides further insight about this dispute, and reveals that the female reference group (blue crosses top right panel of Fig C) of rats fed non-GM diets were super-rats.


Quote from HCB report text translated using Google translate:
The lack of statistical power due to a very low number of control groups, prohibits formally conclude to the presence or absence of an effect of diet on mortality, especially in females. This lack of power can be compensated by the introduction of a priori information on the expected behaviour of the control groups. Of mortality data from the Sprague-Dawley strain used in this study were obtained from the company Harlan (24 ) and can complement the information provided by the study recounted in this article.

The use of the reference data supplied by the producer confirms that it is not possible explain the observed differences in survival curves between experimental groups and control groups by an effect of the test treatment: Figure C shows that the experimental groups are broadly distributed within the prediction intervals. Group Experimental lies at the margin of the interval: about 18 groups, there is nothing wrong with this an observation is outside the prediction interval at 95%. Rates observed survival at two years for the experimental groups are quite compatible with reference data provided by the breeder SD rats. However, we can noted that survival rates observed in the control groups are quite distant from that the reference values for this group suggest (the observed proportion of female rats after two years of life is outside the prediction interval at 95%).

This confirms the fragility of statistical results obtained from such small numbers: it can not extract data from the article no formal conclusion about the effect of different regimes on the survival of rats.