This randomized, double-blind, controlled clinical trial is quite well conducted and remains one of the most robust trials currently available on this treatment. Its pragmatic recruitment approaches real-life effect. Nevertheless, some bias remains:
- Changing inclusion criteria during the trial introducing heterogeneity into the sample...
- Risk of recall bias and classification bias due to the chosen data collection methodology (remote self-questionnaire) and the possibility of death not being found
- The diagnosis of COVID-19 (primary endpoint) may be based solely on the patient's clinical symptomatology collected by questionnaire (no PCR test is required). The criteria defined are rather robust and in the absence of virological confirmation, patients are classified as "probable cases", but there is still a risk of classification bias.
- Follow-up of subjects may not be long enough (measured at D14), with some patients developing symptoms after 3 weeks and not considering asymptomatic cases representing a non-negligible proportion of COVID-19 infections.
- Robust statistical analysis design, including interim analyses, but use of a non-parametric test when a parametric test could certainly have been performed
- Rather female sample of young and healthy patients, not representative of the entire population infected with SARS-CoV-2: decrease in healthy worker possible (majority of healthcare professionals)
- Different patient adherence between groups with lower compliance in subjects taking hydroxychloroquine but caused by adverse treatment effects
bibliovid.org and its content are bibliovid property.