On March 17, Gautret et al. published a manuscript reporting beneficial effects from hydroxychloroquine in COVID-19 in the International Journal of Antimicrobial Agents.1 At a presidential press conference three days later, President Trump discussed hydroxychloroquine, asserting, “It’s shown very encouraging—very, very encouraging early results.” Shortly thereafter, pharmacies around the world reported shortages. Kaiser Permanente went so far as to restrict access to hydroxychloroquine, including for many patients with rheumatic conditions. Rheumatologists everywhere have been left wondering what happened and how to move forward.
At first glance, the enthusiasm for hydroxychloroquine had merit. Multiple in vitro studies have demonstrated encouraging results, and unpublished reports from China suggested a clinical benefit.2,3 Hydroxychloroquine has an acceptable safety profile and a reasonable cost. An editorial in mid-February by Didier Raoult, also the senior author of the Gautret paper, made the audacious claim, “If clinical data confirm the biological results, the novel coronavirus-associated disease will have become one of the simplest and cheapest to treat and prevent among infectious respiratory diseases.”4
Clinical data arrived just four weeks later, when Gautret et al. published their experience treating COVID-19 with hydroxychloroquine in the south of France. The manuscript evaluated 42 patients with PCR-confirmed COVID-19 and reported virological cure in 14 of 20 patients (70%) treated with six days of hydroxychloroquine. Only two of the 14 patients in the control group had a virological cure six days after enrollment. All six patients who also received off-protocol azithromycin cleared the virus. Although small, the results appeared to confirm biological data. A closer read of the manuscript, however, suggests otherwise.
Typical of preliminary data, the trial was small, open-label and non-randomized. Studies of this nature suffer from bias and confounding, and frequently do not translate into successful phase 2 or 3 trials. Whether this study is a “trial,” however, should also be questioned. Although the treatment group was recruited from a single center, the control group was recruited separately from various hospitals in the region. Even more troubling, patients who refused or could not tolerate hydroxychloroquine were included in the control group. This design more closely resembles an unnecessarily biased case-control study than a clinical trial. Its results should have been presented and interpreted as such.
Methodological flaws after recruitment compound problems with interpretation of their data. The primary outcome measure, viral clearance at six days, has not been validated and does not necessarily correlate with outcomes that are clinically relevant, such as mortality. Testing of the control group was less rigorous than testing of the treatment group, and almost half of the control group required imputation of their virus results. Most importantly, six of 26 (23%) patients in the treatment group were lost to follow-up and excluded from subsequent analysis. These included patients with the most relevant clinical outcomes, including adverse events (one patient), ICU transfer (three patients) and death (one patient).