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
More reservations on the Minnesota study claiming small, statistically insignificant benefits from the early hydroxychloroquine treatment. Recall that the primary endpoint of hospitalization was changed midway due to a general absence of severity. Too much self-healing going on.
Why is the rapid self-healing of the young cohort analyzed (median age 40) such a problem? It reduces the opportunity for a therapeutic agent to differentially demonstrate efficacy, as illustrated below:pic.twitter.com/96aJA1gkbk
In this context of a self-healing confounding factor, methodologically, the last thing you want to do it to focus on a final time end point. Tragically, the key results highlighted by the authors do exactly that:pic.twitter.com/BwWcjGeMfg
The study evidences differential convexity and precisely achieves statistical significance at the point of the maximum difference: p=0.05 at Day 10. By nature of the self-healing confounder, you cannot hope to achieve more given the study design.pic.twitter.com/34SUTBdAK8
Achieving statistical significance (p=0.05) is remarkable when 38/165=23% of the HCQ participants did not adhere to the medication protocol (19 pills). Moreover, the self-healing 18-35 year-old only achieved a 5% symptom severity gain gap, but the above 50 year-old subgroup 24%.
There is another serious issue with the mathematical analysis itself, derived from the following disclosure on the Severity Score changes: “the assumption of normal distribution is appropriate”. Well, no. Notice the histograms below and the change in granularity.pic.twitter.com/UHc7nPebp8
Why the change in granularity as the number of patients is not going down? Performing a Shapiro-Wilk W normality test on the Day 14 distribution yields a p=0.0455, just at the point of failure. Were the bottom histograms artificially regrouped to obtain normality? Red flag.
There is a strong argument against normal distribution of score changes due to the study design. The slider below was used for self-assessment of symptom severity. It is range bound. The cohort is self-healing. Score changes will quickly be constrained by the lower limit.pic.twitter.com/t2eANxVUxH
How important is that normal distribution business? Well, if that condition is not met, the statistical analysis of the study is invalid, and the paper should likely be retracted.
Usefully, the study provides data on increasing severity scores, which *might* be less range-bound. It supports the evidence of the convexity phenomenon as well, i.e. the hypothesis of hydroxychloroquine therapeutic activity.pic.twitter.com/Jm6inkT0eE
Another grave reservation on the study: moral hazard. Its self-declarative nature allows liars to join in and distort the numbers. In the hysterically politicized US context, it is only too easy to imagine Trump-haters gate-crashing the party and gleefully reporting fake data.
Cheaters keen to destroy hydroxychloroquine would likely report fake side effects. Compare the side-effect prevalence of the US study with the 3,737 patient cohort of Dr Raoult (similar dosing and age): 43% vs 13%. This is not credible. A huge red flag right here.pic.twitter.com/2SN7pzg9Dq
A cheater does not know to which arm he is allocated, and fake side effect reporting will go up for the control arm too. Dr Raoult reported 0.6% of adverse events in his 162 untreated patients. The US study’s placebo arm? 21.8%. Hypochondriac nonsense or wilful sabotage? Or both?
Nonetheless, I want to express my admiration and gratitude to @boulware_dr and team. Pragmatic clinicians led by scientific investigative open-mindedness, spurned by their government and deprived of funding, urgently explore a drug candidate efficacy and target relevant data. 
Had such a mindset of exploratory curiosity and entrepreneurial courage been adopted by 79-year old Dr. Fauci, bent a reliving his AIDS success story thru novel remdesivir, we would probably be much nearer a solution by now. Remember: 200+ FDA approved drugs kill COVID in vitro.
We need a Marshall plan of systematic drug exploration against COVID-19, broad in scope and amply funded. It would cost a fraction of the countless economic stimulus plans declared to date, but we are simply not getting it.
Conclusion: the first US early hydroxychloroquine treatment study is likely irretrievably flawed, both in methodology, analysis and results. However, it inadvertently points to therapeutic efficacy in the older, at-risk population and more research is badly and urgently needed.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
