Successful republication of Hazan et. al. pioneering work on treating severely hypoxic COVID-19 patients with an ivermectin-based 10-day multidrug protocol.
We are pleased to announce the successful publication of the retrospective case series of 26 hypoxic COVID-19 patients who were seen by Dr. Sabine Hazan in the time period between August 2020 and February 2021, of whom 24 accepted a 10-day treatment protocol that consisted of ivermectin, doxycycline, zinc, Vitamin C, and Vitamin D. A copy of this publication has been deposited on the Scholarworks repository with the UTRGV library. Furthermore, supplementary material (data and computer codes) have been deposited on Figshare.
The patients were treated outside the context of any clinical trial as outpatients via telemedicine. A unique characteristic of the treatment protocol was the dynamically adaptive ivermectin dosage, individualized for each patient, based on the response of room air SpO2 levels to treatment, which required frequent monitoring of the patients during the first few days over 3 to 6 hour intervals until recovery of oxygen saturation levels. The development of the ivermectin-based treatment protocol was pioneered by the senior author of our study, Professor Tom Borody and his research group, who unfortunately passed away on October 6, 2025, after the initial submission of the paper.
Summary of findings
As noted in the paper, the main findings are as follows:
a. At the onset of IDZCT, rapid increase of SpO2 on room air was observed in the 21 hypoxic patients with available SpO2 data on room air for day 2, of which 18 out of 21 were stabilized at SpO2 > 90% within 24 hours;
b. this rapid increase in SpO2 levels was statistically significant for these 21 patients, it remains statistically significant when missing day 2 data is imputed with the baseline SpO2 levels on room air and/or when patients that took or may have taken concomitant medications are excluded;
c. the increase in SpO2 levels was intensified on the severely hypoxic patients that received the 36 mg ivermectin stat dose in addition to the adaptive ivermectin dosage used on all other patients;
d. hospitalization was prevented for all 24 patients accepting IDZCT for a period of 10 days with complete and sustained recovery of oxygen levels on room air by day 10;
e. of these 24 patients, complete resolution of all other symptoms was achieved within 20 days from the onset of treatment for 23 out of 24 patients;
f. All 24 patients accepting treatment survived.
The study does have limitations that you can read about in the paper itself. They mostly boil down to the fact that one cannot make any inductive inferences about treatment efficacy, although we argue that when the findings are viewed in the context of all other research in this area, one can make a favorable Inference to the Best Explanation (IBE) about treatment efficacy. There are philosophical debates about whether or not an IBE is only hypothesis generating and whether it can also establish treatment efficacy in some circumstances. These debates are logically equivalent to having philosophical debates about refining the precise definition of the word “establish” when making claims about “establishing efficacy”. For this paper, we have carefully distinguished between inductive inference and IBE, allowing readers to bring in their own philosophical framework and interpretations.
Further context concerning this publication
These results were first published in a groundbreaking paper in Future Microbiology which was later unethically retracted by the journal on January 9, 2025 with insufficient and non-transparent explanation and in violation of the COPE guidelines. This was a highly destructive action of academic vandalism, as these results were factual, valid, successfully replicated, and very important for their impact in subsequent research. Indeed, a significant body of research literature, reviewed by Scheim and colleagues, relied on the factual report of this patient case series, in addition to my critical appraisal of this ivermectin-based 10-day treatment protocol. As fate would have it, the retraction occurred just a few days after the acceptance of the critical appraisal papers by The Japanese Journal of Antibiotics (JJA). I immediately notified the JJA journal editors, asking for guidance and advice, and simultaneously, I contacted Dr. Hazan’s group offering to independently re-analyze their data from scratch and republish their paper.
A first preprint of the new Hazan case series paper was ready by the time we completed the production process of the critical appraisal papers. The preprint was then submitted for publication to the American Journal of Infectious Diseases, it was sent to 3 reviewers, underwent two extensive major revisions, and is now finally back on the scientific record. The review process was very rigorous. The referees were very detail-oriented about requesting further clarifications and further statistical analysis and we are deeply grateful for their meticulous review of our paper.
The new paper improved upon the original publication in several ways:
1. We explained in further detail the use of adaptive ivermectin dosage.
2. We made the underlying data available on Figshare. Almost all of the relevant patient-level data was already available in the original publication, except for the comorbidity data. Now, all data needed to reproduce our analysis is publicly available, including the computer code that we have used to automatically generate all the tables and figures in our paper directly from the data using the LaTeX typesetting system.
3. We have clarified the setting in which the patients were treated and the process of how they were selected from the general population. We have also provided additional patient-level data about the selection process for these patients.
4. We reported patient-level data about the concomitant use of other medications.
5. We added a complete accounting of the comorbidity counts for all patients. The original publication was very selective about the reporting of comorbidities. The entirety of the available comorbidity data shows that all patients had at least one comorbidity and at least one COVID-19 susceptible comorbidity. The latter were characterized using recent CDC guidelines, which may have been revised by the CDC since the original publication. Readers can repeat the analysis using different characterizations of COVID-19 susceptible comorbidities, by modifying our computer code accordingly.
6. Reviewers requested some additional statistical analysis, as they were very intrigued by the results, specifically about whether any of the known patient characteristics were predictors for time to recovery. With a small sample size, this did not lead to any noteworthy findings, but it was worthwhile to find out if there was anything interesting down that rabbit hole. Reviewers also requested more visualization of oxygen saturation recovery and additional sensitivity analysis.
Predatory retraction by Future Microbiology
Several remarks are in order about the unjustified retraction of the original publication. The retraction note states that:
“Following publication, concerns were raised regarding the methodology used in the study, the patient enrolment process and data analysis. In accordance with our policies, we contacted the authors for an explanation and engaged an independent reviewer. Upon review of all materials, we were unable to confirm the integrity of the study design and the reported findings.”
This is hardly a sufficient explanation for justifying the retraction. We note that:
1. There was no “enrollment process”, in the sense of a clinical trial, because this was a patient case series in which the patients were treated outside the context of any clinical trial as outpatients. This was clear enough by reading the Method section of the original publication, where it was stated that the patients in this case series were “deemed too sick to enter a placebo-controlled trial”. This implies that these patients’ treatment was outside the context of any clinical trial. That narrows it down to off-label individualized care outside the context of any clinical trial. In the new publication we provided further clarifications about the setting of the study. However, these additional clarifications are not particularly relevant to the main results of the paper, and any questions about the setting of the study should have been addressed with a corrigendum rather than a retraction.
2. The patient-level data about the case series that was reported in the original publication was reported correctly, which I was able to confirm independently from the same de-identified data file that was originally extracted from the patient charts and used to prepare the original article.
3. The data analysis in the original publication involved comparing the case series mortality and hospitalization outcomes against historical control groups extracted from a CDC database. This is the correct methodology for investigating an, otherwise, uncontrolled case series. In the critical appraisal papers, we independently replicated this analysis and confirmed the approximately 14% CFR for hospitalized patients in the historical control obtained in the original publication, despite using different methodology (we did not filter for comorbidities) and a different version of the database (there are some variations with respect to missing data across different versions of the CDC database). However, we also found that the 14% CFR was actually a lower bound, because of missing data in the CDC database, and was calculated conservatively by assuming survival for the patients with missing data; thus, the actual mortality rate could have been larger than that. An advantage of using a population-level historical control group is that there is no selection bias in the control group, because the entire population is included. This makes it possible to compare a treated patient case series against the population control group, when the population is either equivalent or at lower-risk than the cohort in the case series, with some caveats about correctly interpreting the results. This argument was explained and developed in further detail on the critical appraisal papers.
4. A previous letter to the editor critiqued the mortality reduction data analysis of the original report by questioning the appropriateness of using an age > 50 and comorbidities > 0 filter in the construction of the historical control group from the CDC database. However, our critical appraisal papers noted that: (a) The age > 50 restriction, used on the original historical control, is an equivalent proxy for the patient’s hypoxic baseline presentation based on previous research on predictive models of inpatient mortality rates; (b) Removing the comorbidities restriction from the external control had a negligible effect on the historical control’s CFR, since we were able to reproduce almost the same CFR despite removing the comorbidities filter; (c) Combining the Hazan and Stone case series provides a far more robust and decisive argument allowing the removal of the age > 50 restriction from the CDC historical controls.
5. The original publication also included a historical control for estimating the patient hospitalization rate. The hospitalization control group (with 30% hospitalization rate), constructed in the original publication, was very conservative, because almost all patients were severely hypoxic and would have been counterfactually hospitalized under standard protocols. Consequently, although the hospitalization control was severely biased towards the null, drawing a conclusion of hospitalization reduction efficacy from a positive finding that survives the bias to null, is methodologically correct. Of course, one cannot quantify the effect, but one can certainly claim that the effect exists. That said, the historical control for hospitalization rates is actually irrelevant, because all patients in the case series were referred to hospital care, which they refused, so the same claim could have been supported more simply with a self-controlled argument.
6. The original publication proposed two patient severity scales, motivated by clinical experience, and some preliminary validation of the scale by comparison against the available time to recovery data. This was only an interesting proposal of patient severity scales, with some initial validation evidence for motivating further investigation, for future researchers to build on this proposal and develop it further. There was nothing controversial about opening this new line of inquiry.
The fact of the matter is that the original report was a factual case series report of hypoxic COVID-19 patients who were successfully treated with a 10-day ivermectin-based multidrug treatment. The reported patient-level data was reported accurately, which I was able to confirm with my independent reanalysis. This data should have never been removed from the academic record. The “methodology” and “study design” was the factual reporting of what happened with these treated hypoxic COVID-19 patients, so I don’t see how the journal could credibly question the “integrity of the study design and the reported findings”. As for the data analysis, the dataset used for the historical controls was publicly available, so the proper way to address any concerns was to conduct an independent critical appraisal by re-analyzing the CDC data. The purpose of scientific publication is to promote scientific dialog. Retracting the paper and removing the factual data of the case series, was an attempt to shut down the further scientific discussion of these findings and vandalize the publication record. It was also an attempt whose failure is now complete.
Conclusion
Dr. Sabine Hazan was recently invited to testify about the unjustified predatory retractions of valid research by corrupted journals and/or publishers in the Permanent Subcommittee on Investigations of the U.S. Senate Committee on Homeland Security & Governmental Affairs by Senator Ron Johnson. The successful republication of this very important data has allowed us to provide further details, insights, analysis with increased methodological transparency. We are deeply grateful to the anonymous external reviewers for very helpful insights for improving the paper and to the handling editor and the American Journal of Infectious Diseases for very high-quality peer review and editorial processes.
References
1. S. Hazan, A.C. Vidal, E. Gkioulekas, A.W. Gunaratne, S. Dolai, R.L. Clancy, P.A. McCullough, T.J. Borody: “Rapid recovery of peripheral oxygen saturation in hypoxic COVID-19 patients with ivermectin/doxycycline/zinc multidrug therapy“, American Journal of Infectious Diseases 22 (2026), 4-26
2. E. Gkioulekas: “Data and materials: Rapid recovery of peripheral oxygen saturation in hypoxic COVID-19 patients with ivermectin/doxycycline/zinc multidrug therapy“. figshare. Online resource. (2026), https://doi.org/10.6084/m9.figshare.30929246
3. E. Gkioulekas, P.A. McCullough, C. Aldous: “Critical appraisal of multi-drug therapy in the ambulatory management of patients with COVID-19 and hypoxemia. Part I. Evidence supporting the strength of association“, The Japanese Journal of Antibiotics 78(1) (2025), 2-34
4. E. Gkioulekas, P.A. McCullough, C. Aldous: “Critical appraisal of multi-drug therapy in the ambulatory management of patients with COVID-19 and hypoxemia. Part II: Causal inference using the Bradford Hill criteria“, The Japanese Journal of Antibiotics 78(1) (2025), 35-68
5. D.E. Scheim, P.I. Parry, D.J. Rabbolini, C. Aldous, M. Yagisawa, R. Clancy, T.J. Borody, W.E. Hoy, “Back to the Basics of SARS-CoV-2 Biochemistry: Microvascular Occlusive Glycan Bindings Govern Its Morbidities and Inform Therapeutic Responses”, Viruses 16 (2024), 647.
