Critique of the study (not a rigorous scientific write-up, not an MD):

Mehra,2020 study:

https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)31180-6/fulltext

On cursory review of the study above (Mehra, 2020) it is very a solid work which demonstrates a very surprising  result not only that hydroxychloroquine (HCQ) has no effect on Covid-19 infection for hospitalized patients, but that it also results in very significant mortality for those patients.

That is a very surprising result for medication which was deemed safe for decades to be used by Lupus among others auto-immune disease sufferers  and Malaria prophylaxis. To be more specific (Mehra, 2020) shows Mortality of 18% for patients treated with HCQ vs 9.3% for patients not given the drug, this is doubling of mortality, how can something like that happen for a drug known as generally safe?

In fact this goes against what was found in (Geleris, 2020) https://www.nejm.org/doi/full/10.1056/NEJMoa2012410

The results of the above study showed that HCQ treatment produced no benefit, but also produced no harm. Geleris is also an observation study, just like Mehra, not a randomized control study, and also reviews the use of HCQ on hospitalized patients.

How would Mehra explain the significant patient mortality in his study when patients are given HCQ?

“Chloroquine and hydroxychloroquine are associated with concerns of cardiovascular toxicity, particularly because of their known relationship with electrical instability, characterised by QT interval prolongation (the time taken for ventricular depolarisation and repolarisation). ”

“However, these drugs prolong the QT interval and increase the risk of sudden cardiac death.”

In addition to increased mortality, the possibility of which was being explored by the Mehra study, something else was also discovered: almost tripling the use of mechanical ventilation for HCQ group versus Non-HCQ group. Now, HCQ and or Macrolides have a known effect on heart QT interval, but nothing is mentioned about the effects of those drugs on the respiratory function. This is our first warning signal, that something could be wrong with the data.

In bellow cut out we can see Ventilator use for HCQ is 20.4% vs 7.7% for no-drug group

At this point we need to discuss the differences between the Randomized Control Trial (RTC) vs the Observational study such as Mehra or Geleris. In RTC we randomly give either Treatment or a Placebo to 2 groups of patients. The patients who get either the control or placebo don’t know what they got, so they can change their behavior accordingly. By giving the treatment or a placebo randomly to patients we remove bias, such as giving treatment to sicker patients. The main issues with RTC comes with difficulty of creating the study, and of ethical dilemmas when it comes to treatment of a potentially deadly disease like covid-19.  That is: if the treatment is found to work, giving the patient placebo potentially sentenced some of them to death.

To potentially help as many patients as possible in critical care the real medication is given to the patient not the placebo. In that case we have no control group, and have no way to compare the efficacy and safety of the treatment. However, what we can do is create a Synthetic Control Group - patients who had the disease, didn’t receive the treatment by some reason, but who match the treatment group in confounders: such as having same health status, symptoms, age, sex - to do that, researchers use “propensity score matching”.

A very simple example of that would be as follows: say we have Bob who happens to be sick with covid-19, who was treated with HCQ, Bob on admission also had fever of 100F, spO2 of 85%, 65 years old and a Male. To match him to a control who didn’t get the HCQ treatment we need to find some other person who happens to also have Covid-19 and as close to Bob’s medical condition, that is also have 100F fever be 65 years old male, etc.. it becomes clear that the quality of our matching is directly related to the number and quality of relevant medical confounders - if we have too little we will be matching people who are clinically very different. If in the above example we drop fever, we might match Bob with someone who happens to have covid19 but who also only has 98.6F body temp, obviously those 2 individuals are clinically different and yet they would be matched up, if we are not careful.

In fact this appears to be what is happening in the Mehra study, the study only looks at 2 baseline vital signs demonstrating the disease severity, a very coarse  spo2 measure of either above or below 94% and qSOFA, on another hand Geleris is looking at 6 vital signs,and an spo2 as a continuous variable instead of as a binary - which it is not.

Geleris:

Even  though Mehra appears to be very well matched - its is not, when it comes determining of covid19 infection clinician would look at many more vital signs and symptoms than spO2, but the paper doesn’t mention much more than that.

Mehra:

What we see above is baseline disease severity based on scarce indicators and the ones which we even have are vague and low fidelity this results in failure to match on critical vital signs between control and treatment groups. Despite Mehra’s best efforts, the control patient subset had less severe disease than the HCQ group - that would make sense, since the most severe patients are being treated with experimental medications like HCQ.

In fact the percentage of mortality in Geleris matches with Mehra, when Galeris doesn’t account for confounding factors.

Geleris:

.

Geleris Intubation or death for HCQ is 32.3%, with Mehra intubation or death is  29.1%

Without HCQ for Geleris 14.9% and Mehra Intubation or Death is 13.2%

But with propensity score matching in Geleris this difference disappears

In Mehra the HR for HCQ is 1.335, while in Geleris the HR is around 1 - demonstrating neutral effect.

“In the primary multivariable analysis with inverse probability weighting according to the propensity score, there was no significant association between hydroxychloroquine use and the composite primary end point (hazard ratio, 1.04; 95% CI, 0.82 to 1.32) (Figure 2). There was also no significant association between treatment with azithromycin and the composite end point (hazard ratio, 1.03; 95% CI, 0.81 to 1.31).”

To underline the problem with just using spO2 measure for covid19 patients we can see  Table 2, Mehra that only 8.4% patients had spO2 below normal in survival column - that is the other 91% had normal “sats” that raises the question of why were they admitted into the hospital to begin with, for the deceased column 80% of patients had normal spO2 on admission. It's clear that they had other serious symptoms of the disease to justify admission into the hospital, yet those symptoms are not included in the paper.

The hint that we were looking at significantly more sick individuals in the HCQ group was the significant increase in Mechanical Ventilation for those patients which could not be explained by QT prolongation.

In yet another recent study out of New York Hospitals, published in JAMA (Rosenberg, 2020)

https://jamanetwork.com/journals/jama/fullarticle/2766117

We see the following indicators of disease severity, significantly larger gamut than in Mehra

“In the primary analysis, following adjustment for demographics, specific hospital, preexisting conditions, and illness severity, no significant differences in mortality were found between patients receiving hydroxychloroquine + azithromycin (adjusted HR, 1.35 [95% CI, 0.76-2.40]), hydroxychloroquine alone (adjusted HR, 1.08 [95% CI, 0.63-1.85]), or azithromycin alone (adjusted HR, 0.56 [95% CI, 0.26-1.21]), compared with neither drug (Table 3) (complete case analysis variable completeness was 86%).”

Conclusion:

1. Based on other studies HCQ shows no efficacy with macrolides or without in critical hospitalized patients.
2. Despite this poorly analized study, it doesn’t kill patients, it just has no beneficial effect thus giving it at the later stage of the disease is pointless.

3. Precisely because of the ease with which it is possible to produce statistical mistakes in Observation studies we have no choice, despite ethical consideration to have a proper RCT for non critical patients: With at-least 2 arms one placebo another HCQ+Zinc and Macrolide

FkretreGtrkc687324

HCQ_LANCENT_CRITIQUE_DEATH_2020