The results of clinical trials are making their way to publication showing hydroxychloroquine to be effective in many, although possibly not all, situations. The American media is silent on these studies. Instead, the American media headline a non-experimental and biased study showing the drug to be ineffective and, even, a cause of death.
The non-experimental study concerns a large number of COVID-19 patients (sample size 368) at Veterans Administration medical centers. These were older men, many with comorbidities. The study is non-experimental because patients weren’t randomly assigned to control and experimental groups. Instead, the sicker patients tended to be administered hydroxychloroquine. It is recognized by the authors of this study that the correlation of severity with prescription imparts bias into the sample. The authors reduce (but do not eliminate) this bias through “propensity score adjustment.”
LINK TO VA STUDY:
When asked about this study on Tuesday evening at the President’s daily briefing on the coronavirus pandemic, Stephan Hahn, Commissioner of the FDA, pointed out that the study was retrospective (not a controlled experiment) but nevertheless informative to those who prescribe drugs to patients. He also said there were thirty clinical trials underway that will possibly provide definitive information regarding efficacy and safety.
Actually, we already have the results of some clinical trials.
In a French study, a small sample of patients entering the Méditerranée Infection University Hospital Institute in Marseille testing positive for COVID-19, with certain exclusions, were recruited to be the test subjects in the study. Other patients elsewhere in the country served as the control subjects. The focus of this study was on patients who were asymptomatic or had mild symptoms. While the sample was small, the drug was shown to be statistically significant in reducing negative outcomes, which effects were improved with the use of Azithromycin.
LINK TO FRENCH STUDY:
In a very small follow-up study of patients (sample size 11) with severe symptomatology, a second French study found that hydroxychloroquine did not help.
LINK TO SECOND FRENCH STUDY:
In a Chinese study that has not yet undergone peer review, 62 COVID-19 patients entering Renmin Hospital of Wuhan University were randomly assigned to control and experimental groups. In this study, hydroxychloroquine was found to help patients fight off pneumonia.
LINK TO CHINESE STUDY:
A second Chinese study, involving 100 patients at multiple locations, obtained essentially the same results.
LINK TO SECOND CHINESE STUDY:
We economists sometimes envy our counterparts in the so-called hard sciences, for their ability to conduct controlled experiments. Almost all of our statistical work is with “retrospective” data. Separating causation from correlation is sometimes impossible with non-experimental data; and, is never easy. We have developed some techniques to attempt to do this, but I’ll not go further into this matter here. I’ll just say that the medical sciences, anticipating the arrival of definitive information from controlled experiments, is not being unreasonable in merely reducing bias through techniques such as propensity score adjustment, when conducting retrospective studies.
Having said this, when speaking to persons outside the field, it is important to clarify that your findings are merely a lower limit as to the efficacy of hydroxychloroquine. If the finding was that hydroxychloroquine improved patient outcomes, then we’d know something. We’d know that the drug is effective, although – because of the bias – we wouldn’t know by how much. As it is, all we know from the VA study is that hydroxychloroquine might not be effective