In a journal club with
@harvardmed
students last week we talked about this issue — death can’t be a censoring event in an analysis of clinical improvement.
These letters all raise points that should have made
@NEJM
think twice before publishing an industry-funded case series.
So, apparently, the statistics in the NEJM report on compassionate use remdesivir without a control group were wrong. 53 authors and not one knows how to conduct a Kaplan-Meier analysis??
@enricoferroMD
@harvardmed
@NEJM
The issue here is censoring vs competing risk.
Censoring must be non-informative: we don’t know if pt will experience outcome later (e.g. lost to f/u)
Competing risk is a different event that precludes the outcome (e.g. death)
@bnrome
@harvardmed
@NEJM
Yes, but bad or questionable research sometimes gives others ideas for research that produces valuable results. Spine pain is a good example. A study of the
#fibromyalgia
#syndrome
exam on pre and post anterior c/spine fusion patients might be of interest to spine pain patients!
@bnrome
@harvardmed
@NEJM
Removing dead patients from the final analysis seems to be a common pattern with industry-funded studies. It’s almost like they’re trying to deny that what they’re trying to sell kills people.