Following a previous post about the NEJM decolonisation paper, there was lots of insightful commentary from around the globe.
As a result, we’ve proven the stats were wrong and an erratum will be published.
Perhaps more significantly, this conversation has shown the true power of this form of peer review, having genuine impact on big name practice changing clinical trials.
Thank you to everyone who has been involved in the discussion, particularly Gordon Doig, Sandra Ware (Research manager, GSA-HEMS), and Simon Carley and for the prompt and gracious reply from the paper’s lead author, the highly accomplished Susan Huang.
Read more to find out the the real stats and how to put these into context for your practice…
In case you missed the orginal discussion, the paper had claimed that decolonisation with twice-daily intranasal mupirocin for 5 days and daily bathing with chlorhexidine-impregnated cloths for the entire ICU stay could have significant implications. In particular, that 54 patients would need to undergo decolonization to prevent one bloodstream infection from any pathogen.
However the numbers given in the paper didn’t correlate with this Number-Needed to Treat (NNT)
Quick recap: the NNT = 1 / (Absolute Risk Reduction)
As Sandra pointed out:
You need to use patient numbers to get number of patients
Relevant numerators are found in table 3
Relevant denominators are found in table 1
For group 3:
MRSA clinical culture
Baseline = 240/17356 = 1.38%
Intervention = 217/26024 = 0.83%
Difference = 0.55% NNT = 100/0.55 = 181.5 close to the 181 reported
But…
Blood stream any pathogen
Baseline = 412/17356 = 2.38%
Intervention = 356/26024 = 1.37%
Difference = 1.01% NNT = 100/1.01 = 99.4 not close to the 54 reported
Prof Huang agrees with this calculation, and the error occurred as the published NNT was derrived from a previous version of the data.
So, significantly, you would have to treat 99 patients to prevent one blood stream infection, and if you look at their mean lengths of stay, that is 388 patient days of treatment to prevent one blood stream infection.