The solid lines supposedly show the relationship between the change in bone turnover (NTX) and new spine fractures in patients taking Actonel (Risedronate) at 1 year and at 3 years. The dashed lines show patients taking placebo. The purpose of this posting is simply to show that the scale of graphs in ghost-written papers by P&G had been drawn so that a proportion of the data simply "fell off" the left hand end of plots. I was supposed to have signed off on two further publications containing similar graphs and opaque statistical analysis ...... all while being refused access to underlying data codes held by P&G.The first step of any statistical analysis is to PLOT THE DATA. Following press exposure P&G eventually released the raw data to authors in April 2006. The distribution of the actual data for NTX change (the X axis variable) in the above paper (Eastell et al 2003) for patients taking Actonel is:
A third of the data would not have appeared within the range of the Procter and Gamble graph as plotted by them (P&G's graph ends at -60%). Strange that.The relevance of the missing end of the plot will be discussed in later postings. Suffice to say that these patients show responses more typical of the drug produced by P&G's competitor (Merck, Fosamax).
Earlier|Later|Main Page
2 comments:
I'm not making out what your University did about this.
Universities do not have the balls to deal with commercial concerns disinterestedly, but it is still their responsibility to make sure that academics and academic freedom are protected. Is there any report showing what attempts the University of Sheffield made to urge P&G to supply the codes?
No doubt you are going to tell us that the rate of V fractures was lower in those off the graph than for those on the graph. I not surprised.
I read the correspondence with the JBMR and am dumfounded. Why don't you submit this work somewhere else instead of blogging? A shameful and embarrassing affair.
Post a Comment