Study: Outcomes of hydroxychloroquine usage in United States veterans hospitalized with Covid-19

We continue from (CNN) … is Wrong in so Many Ways About Hydroxychloroquine…the Facts Part 2 with this brief note.

The study of the moment is (medRxiv) Outcomes of hydroxychloroquine usage in United States veterans hospitalized with Covid-19.

The popular maelstrom picks up on something like this with proclamations of worthlessness and political finger pointing. The value of this study is too small to become part of that. It usefully warns and reminds the clinician that:

  • Hydroxychloroquine is a dangerous drug.
  • If it does work, there is likely to be a tight regime, with the narrow therapeutic index characteristic of drugs of the pre-antibiotic era. With some of these drugs, the range between the effective dose and the lethal dose was exceedingly small.
  • No one has any idea how to dose. Experimental dosage is based on toxicity to the malaria parasite, entirely different from the job of chaperoning zinc ions into an epithelial cell.

Quoting,

In this study, we found no evidence that use of hydroxychloroquine, either with or without azithromycin, reduced the risk of mechanical ventilation in patients hospitalized with Covid-19.

The words are correct, but should not be confused with “found evidence that [the combinations] did not reduce the risk of mechanical ventilation in patients .”  The study found nothing. either pro or con.

The link contains reader comments, which are well thought out. One factoid draws attention. Quoting,

Rates of ventilation in the HC, HC+AZ, and no HC groups were 13.3%, 6.9%, 14.1%, respectively.

This is a red flag. The ventilation rate for HC+AZ recipients was half that of the others. There are two ways to take this:

  • The HC+AZ recipients did not enjoy a statistically meaningful reduction in mortality, so, so what?
  • Something is wrong with the data.

Whenever we analyze statistical data, we look for “outliers”, data points are not coherent with the trend. The 6.9% could be the result of outliers in a too-small, invalid trial, or it could represent a unidentified cohort with salient characteristics.

Three choices are available:

  • Continue the study until it is large enough that the salient characteristics of the cohort can be identified.
  • Decide that no cohort exists. Exclude messy data as outliers.
  • Decide that the data does not reach the threshold at which statistics can be applied.

This study does not reach the threshold of statistical validity, and should not become part of public policy debate. It has value only as a caution to the clinician.