Correction to “Optimizing Hydroxychloroquine Dosing for Patients With COVID-19: An Integrative Modeling Approach for Effective Drug Repurposing”

Abstract

Garcia-Cremades, M., Solans, B.P., Hughes, E., Ernest, J.P., Wallender, E., Aweeka, F., Luetkemeyer, A.F., & Savic, R.M. Optimizing hydroxychloroquine dosing for patients with COVID-19: an integrative modeling approach for effective drug repurposing. Clin. Pharmacol. Ther. 108, 253–263 (2020). https://doi.org/10.1002/cpt.1856.

Following the retraction of the publication “Hydroxychloroquine and azithromycin as a treatment of COVID-19: results of an open-label non-randomized clinical trial” by Gautret et al. [1], we write to report a correction to our article in Clinical Pharmacology and Therapeutics, “Optimizing hydroxychloroquine dosing for patients with COVID-19: An integrative modeling approach for effective drug repurposing” [2]. The Gautret et al. [1] publication was retracted on the basis that methodological flaws were identified, including issues with study design, data handling, and statistical analysis, which ultimately compromised the validity and reliability of the results. Given that our article relied in part on the findings of Gautret et al. [1], we have carefully reviewed our analyses to ensure the integrity and accuracy of our conclusions.

Our publication synthesized a comprehensive body of knowledge to develop model-informed dosing recommendations for hydroxychloroquine. We integrated emerging data from preclinical evaluations and in vitro antiviral testing—largely derived from early COVID-19 studies—with the extensive pharmacological knowledge accumulated over decades of hydroxychloroquine use in malaria. This included published clinical population pharmacokinetic models, exposure-efficacy relationships, and exposure-safety data. We used two independent studies to compare results, evaluate the variability of COVID-19 natural history and effect on drug efficacy, and to validate our modeling outcomes. By integrating this totality of evidence, we were able to propose informed dosing regimens tailored to the COVID-19 context. Our analyses determined that hydroxychloroquine doses of 400 mg or below twice daily for five or more days were predicted to have no effect on viral loads and reduction of the proportion of patients with detectable SARS-CoV-2 infection. However, we also found that doses exceeding 600 mg twice daily were predicted to result in clinically concerning QTc prolongation. We acknowledged that this finding had potential safety implications that would require careful prospective assessment.

We estimated a clinical EC50 value of 5.3 μM using data from the Gautret et al publication to simulate outcomes in Figure 6 [2] The in vivo EC50 value was in the range of the in vitro EC50s reported in Figure 4. The geometric mean of the in vitro EC50s was only slightly higher (9.95 μM versus 5.3 μM), and if we replaced this value in the simulations of Figure 6, we would come to similar conclusions.

A later trial done in Brazil tested 400 mg twice daily and found that it was indistinguishable from controls and that patients receiving hydroxychloroquine had higher QTc prolongation than those on non-hydroxychloroquine containing regimens [3]. These results are in line with our prediction that doses greater than 400 mg twice daily would be necessary to observe an effect. Furthermore, subsequent studies have found that baseline QTc prolongation was increased in patients with COVID-19 independent of treatment [4, 5]. These findings were among data that emerged after our publication and could have further informed our safety assessment. The modeling results provided an early, actionable quantitative assessment of the clinical pharmacology of hydroxychloroquine to guide further trials. Despite retraction of the data used from one publication, the totality of the data integrated and modeling was capable of predicting trials that would confirm the lack of efficacy of hydroxychloroquine in patients with COVID-19.

The value of our work lies in demonstrating how a focused and coordinated effort, integrating all available information at unprecedented speed, can meaningfully inform clinical decision-making. Our team—bringing together clinicians, clinical and translational pharmacologists, and modelers—worked with urgency and precision to provide timely guidance on potential dosing strategies for hydroxychloroquine, based on the best scientific knowledge available at the time.

A key insight from our analysis was the identification of a very narrow therapeutic window—suggesting that doses potentially effective for antiviral activity were also associated with a significant risk of toxicity. This concern, raised early in our modeling work, was subsequently confirmed by several prospective clinical trials. Our goal was not only to flag these critical safety considerations but also to propose rational dosing strategies that could be evaluated in a formal clinical setting. Indeed, the regimens we recommended were intended to inform a prospective NIH-funded clinical trial.

While one specific data source—the study in question—was later retracted, this occurred several years after our analysis. At the time, this information was not known to us, and we acted in good faith, relying on what was then considered credible scientific evidence published in a peer-reviewed journal. Importantly, this particular study constituted only a small component of the broader evidence base we integrated, and its removal would not have meaningfully changed our overall conclusions.

Our aim was to demonstrate the power of modeling and simulation approaches to respond rapidly and responsibly to emerging public health threats, offering data-driven guidance to support rational and evidence-based decision-making. The team came together in mid-March 2020, and within just 10 days, our group of seven clinicians, pharmacologists, and modelers completed the analysis and manuscript.

To provide context, human-to-human transmission was confirmed in China and the first US case was reported in Washington State on January 20, 2020. Our manuscript was submitted on March 31, 2020, and accepted by April 12, 2020. This timeline underscores the vital role that rapid, collaborative modeling efforts can play in addressing global health emergencies.