That was fast.

So overnight, the data used in the Lancet publication on hydroxychloroquine has come under serious question.  In short, reports in UK media this morning (here at least) that Australia happened to notice that the deaths attributed to Australia in the study during the time period studied were -higher- than what Australia actually had.  The researchers claim this was a mistake including a hospital actually located in Asia as Australian.  That would be understandable for a study that large, but ideally would not happen.

Based on this, epidemiologists in Australia have asked for the participating hospitals in Australia to be named.  Some of them (including some very large and well known hospitals there) have not heard of the database that was doing all the data collection.  Now, that’s not impossible–there are a number of “real world evidence” vacuums like this, and it’s tough to keep track of them all.  So far, the database in question (Surgisphere) has refused the call to make its data public though, which is not a great sign. 

Elsewhere, some statisticians are quibbling with some of the methods in the paper, but that tends to happen too.

That said, neither the Australia discrepancy nor the statistics quibbling are apparently enough on their own to alter the study conclusions.  If they are evidence of wider problems in that data set though, look out.

So we’ll see.  The trials on hold around the globe for hydroxychloroquine should be making an independent go/no go call based on the data they have collected so far anyways.  If there is no safety signal (i.e. hydroxychloroquine is NOT obviously dangerous in their study), and no futility signal (i.e. hydroxychloroquine is NOT likely to separate from the control arm), those studies will likely resume.  Meanwhile, the Lancet paper authors will be pretty busy trying to straighten out the database they used, and expect at -least- a correction to that paper.

<Paladin>