Alex Berenson has published a gloating article citing a new study – published in JAMA (the Journal of the American Medical Association) – claiming to show that ivermectin has no significant effect on reducing severe disease progression in high risk patients with Covid (and note the gratuitous dig at Dr Robert Malone).
Here is the graphical abstract of the JAMA article describing the study with the ludicrous use of donut pie charts to compare percentages to summarise their ‘main result’:
But it turns out that the JAMA article is yet another example of what can only be called scientific publishing fraud (similar stunts were used to discredit the use of hydroxychloroquine as explained here). The main paper obfuscates the results of the outcome we are most interested in: whether those given ivermectin were less likely to die than those in the control group.
As pointed out on twitter by Jikky Kjj and Massimaux in a tweet reply to Pierre Kory (and in this article by Jessica Rose) only 3 the 241 patients given ivermectin (1.24%) died compared to 10 of the 249 patients in the control group (4.02%).
Using the standard (but useless and arbitrary) ‘p-value’ approach to statistical significance, this falls just under the level required to claim the result is ‘significant’. This mortality information is crudely buried inside a paragraph on page 10 of 12 of the paper and is not mentioned in the abstract, results, or conclusions:
Note the ridiculous statement “The ..mortality rate was similar for the ivermectin and control groups“. In fact, there’s nothing ‘similar’ about them at all. A Bayesian analysis of the data** (assuming a uniform prior assumption for probability of death in each of the ivermectin and control groups) results in a risk ratio with median value 0.356 and 95% range 0.096 to 1.046. The probability that the risk ratio is less than 1 is just under 97% (it is 96.88%). In lay terms you can interpret this as a 97% probability that the true death rate for patients taking ivermectin is less than that of those who don’t. And playing the game of ludicrous donut pie charts, here is what one for death (that they didn’t include) looks like:
While the paper contains many tables with some bizarrely unnecessary information, the crucial mortality details (except for the obfuscated statement) are not in the main paper at all. You have to go online and find Supplement 2 and look for eTable 6*:
Unlike in the main paper, this table lays out the mortality data clearly, and also includes mortality by vaccination status. Why was this not included? Moreover this supplement, as well as the full article, do not appear to be ‘open access’ – I could only access them through a university subscription. Only the abstract, with its misleading results, is generally accessible. So people like Alex Berenson (and others who seek to discredit early treatments for Covid) are able to widely publicise misleading information to an unsuspecting public who are generally unable to see the real results for themselves.
JAMA need to be held to account for publishing the article in this form – deliberately hiding the most important results.