You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
Badly flawed large (n=5,683) database analysis from Peru studies various treatments with hydroxychloroquine +/-azithromycin+/-ivermectin and concludes that nothing works vs standard of care, or worse.pic.twitter.com/RUQy1Hj1kS
The dataset shows a 9.3% higher mortality for the HCQ arm. Thanks to the randomized RECOVERY trial (see below), we know for sure that HCQ has no negative impact on hospitalized patients' mortality. So the dataset has a critical confounding bias.pic.twitter.com/CLzGci7DPg
Is the critical bias properly redressed? The HCQ data shows a 61% higher mortality risk. The hazard ratio evaluated by the authors after various fanciful statistical analyses ranges from 1.39 to 1.73: there is just zero evidence that the confounder has been eliminated.
The other extraordinary thing is that IVM, AZT+HCQ and IVM+AZT also increase mortality to a similar extent: all hugely lethal, whereas they are all considered very well-known safe drugs. What about a bit of common sense?
The statistical thresholds are adding to the general non-sense of the paper: Ivermectin causes death or ICU admission, but not death nor death or O2 supplementation. What??? Laudably, the authors briefly smell a rat, to only ignore it later.pic.twitter.com/4zJ0OXLsmf
This is just horrible statistics and clinical science. What is going on? It is trivial to anyone with a modicum of medical judgement. You just have to plot this:pic.twitter.com/Gy5K5vBVSG
The simple number of patients treated explains 94% of the variance of the mortality rate! An old classic: either due to scarcity of the drugs and/or compassionate use and/or "unproven" drug risk aversion, HCQ and Ivermectin were restricted to the patients in worst condition.
Azithromycin in Peru seemed to have been more consensual and/or available, therefore it was prescribed to more patients, who were mechanically less risky.
The anomalous prescription pattern and AZT's fairer shake are actually quite visible in the appendix that provides the support assessment for the flawed PSM analysis. If the confounder is not fully captured by your covariates, the analysis is... worthless. Which is were we are.pic.twitter.com/87JFo1fDtM
When you are prejudiced to reach an anti-HCQ/IVM conclusion, you will become remarkably blind to data staring at you and telling you what you do not want to accept. All the HCQ/IVM arms end up outperforming SOC after a while, i.e. saving lives:pic.twitter.com/qqqQAG8bXt
The severity/indication bias is visible on all the KM charts, causing the treatment to underperform initially. The HCQ chart is particularly fabulous with ca. 15% of the patients dying on Day 2: (RECOVERY chart for comparative sanity)pic.twitter.com/0NmZpTs16u
If the first two days are excluded, the conclusion on HCQ vs SOC becomes entirely different:pic.twitter.com/yAfLYOhQTn
15% of deaths on Day 2 makes you appreciate the quality and trustworthiness of the study's method as well, as described by the authors... Garbage In Garbage Outpic.twitter.com/BIufzrILBW
And finally: birds of a feather flock together. The paper makes ample reference to the other recent, analytically-beyond-feeble study from Fiolet. What a surprise... Hopefully the national policy of Peru (or any other country) will not be influenced by such low standard works.
The paper itself, ironically called "Real World Effectiveness". Would be comical if thousands of lives were not at stake.https://www.medrxiv.org/content/10.1101/2020.10.06.20208066v2 …
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
