We want to rapidly know as much as we can about COVID-19. In the current pandemic, researchers are rushing to complete studies and often prematurely release the results on MedRxiv (pronounced “med-archive”) before they have even undergone a process of peer review.
Of course, this can be good news for rapid dissemination of information, but it can also be rather harmful if bad studies get pushed without proper scrutiny.
This multicenter RCT from china included 150 patients with COVID-19. They got hydroxychloroquine (open label) vs. standard care. The primary outcome was clearance of the virus at 28 days.
They concluded, “…adding hydroxychloroquine to the current standard-of-care in patient with COVID-19 does not increase virus response but accelerate the alleviation of clinical symptoms.” (Yes, you read that correctly).
They go on to say that the alleviation of symptoms may due to the anti-inflammatory properties of hydroxychloroquine and less lymphopenia. The authors suggest clinicians might consider it in symptomatic patients with elevated CRP and or lymphopenia because it might prevent disease progression in high risk patients.
Unfortunately, this conclusion is wrong.
Their primary outcome was negative, but at with such small numbers was at risk of a Type I error. Hidden in the manuscript, the authors mention that they wanted to enroll close to 400 patients but had to give up when recruitment fell short.
The median time from symptom onset to getting hydroxychloroquine was 16 days. For most antivirals to work, they need to be given early.
But what about the claim that the medication reduces clinical symptoms?
This comes from one of fourteen post-hoc subgroup analysis in 28 patients that was “statistically significant” with hazard ratio 95%CI of 1.09 to 71.3.
Peer reviewers will easily spot this inappropriate claim that came about through a process known as data dredging or P-hacking.
What should we conclude?
This was an underpowered biased study that made some silly claims based on an unadjusted post-hoc subgroup analysis.
We still don't know if hydroxychloroquine helps or harms.
Tang W, Cao Z, Han M, et al. Hydroxychloroquine in patients with COVID-19: an open-label, randomized, controlled trial. MedRxiv [link to un-peer reviewed manuscript]
Post a Comment