About the Lancet paper on hydroxychloroquine and COVID-19.

Copyright 2020 Robert Clark

 Attached below is an email I sent to the authors of the recent Lancet paper on HCQ and COVID-19:

Hydroxychloroquine or chloroquine with or without a macrolide for treatment of COVID-19: a multinational registry analysis.
Prof Mandeep R Mehra, MD  Sapan S Desai, MD  Prof Frank Ruschitzka, MD  Amit N Patel, MD
Published: May 22, 2020DOI:https://doi.org/10.1016/S0140-6736(20)31180-6
https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)31180-6/fulltext

 First of all, the paper asserted that it only included patients that received the medications within 48 hours of diagnosis. This unfortunately gave the impression that this meant it was a case of "early treatment" with HCQ.

 The proponents of HCQ have argued it must be given early to prevent progression to serious illness, where it is much less effective. In fact, they argue it should be given on first sign of symptoms, even if before a positive test result comes back. The reason is it may take up to a week for the test results to come back, and over that time the virus is getting worse.

 Then this study really wasn’t early treatment despite what was said in the paper for reasons explained here:

James Todaro, MD @JamesTodaroMD May 22

The Lancet study gives the appearance of "early treatment" w/ HCQ, but this is NOT the case.

Symptom onset to hospitalization = 7 days
Hospitalization to diagnosis = 2 days
Diagnosis to treatment = 1-2 days

Time from symptoms to HCQ treatment = 10+ days
https://twitter.com/jamestodaromd/status/1263891253965590529?s=21

  Also, the cases in the Lancet article were those diagnosed from Dec. 2019 to mid-April. During this earlier time, it took several days to get test results back sometimes as much as a week:

https://www.npr.org/sections/health-shots/2020/03/28/822869504/why-it-takes-so-long-to-get-most-covid-19-test-results

 So for most cases in the Lancet study the actual time between symptom onset to treatment was likely greater than two weeks.

 The authors of the Lancet article have acknowledged also the study was only concerned with hospitalized patients, and people with COVID-19 symptoms only require hospitalization when their symptoms have progressed to a severe level.

 For this reason, the authors also have stated the results of the study should not be regarded as applicable to use of HCQ in an early, outpatient treatment format.


 Below, my letter to the authors:

About your article on HCQ and COVID-19.

RC

Robert G Clark

Mon 5/25/2020 10:42 AM

 Hello, Dr. Mehra. I was very interested to read your article on HCQ treatments for COVID-19. The problem with such studies, absent of RCT’s, is the sicker patients get the test drug and the healthier patients do not. This skews the mortality in the test group to be worse and the mortality in the control group to be better.

 I just saw a problem with how you tried to account for that:

Hydroxychloroquine: When medical science starts to look like political science.
May 23, 2020

https://drquay.com/hydroxychloroquine-political-science/

 It discusses the attached table from your report. This shows that the number of HCQ patients on ventilators was over twice the number as the controls. This shows your study didn’t sufficiently match for severity of disease between the two groups. Being on a ventilator is of course is a high indicator of a poor final outcome. Also, for severe cases as well of COVID-19 it is known the disease can causes heart problems, which also explains the higher number of heart problems seen in the HCQ group.

 So tell me if this is a reasonable way to instead approach the question of equalizing the HCQ and control groups. Lets first look at all the patients in toto. I assume the medical records were taken before being assigned the medication. Look at the proportion of cases for each category such as hypertension for all the patients. Then scale up or down the number of cases in the test and control groups for each category so the proportions match those for all the patients.

 Now, take for the number of deaths for each category in the test and control groups to make the mortality ratios match those observed in the original study data. Note this may mean the numbers won’t match what the actual numbers are for each group, but what we are trying to do is equalize the proportions for the test and control groups so we can get a fairer comparison between the groups.

 In this way you have for both the test group and the control group equal proportions of cases with the high risk factors and you have the mortalities estimated from the original data found.

 That’s for a first level analysis. But there is a problem here in that we know for patients that have multiple risk factors their mortality would be worse. But we are looking at the risk factors only one at a time. To get an even more accurate analysis we break up all the cases into further subcategories when patients had two or more of the risk factors. We then make the proportion of cases in the test and control groups match these proportions and again make the mortality ratios in the subcategories match those in the original study.

 So can you provide the raw numbers, before the propensity adjustments, say, as a supplement to your report, so we can calculate these ratios ourselves?

  Thank You,

    Robert Clark

___________________________

Robert Clark

Dept. of Mathematics

Widener University

One University Place

Chester, PA 19013 USA

___________________________