Rapid increase of SpO2 on room air for 34 severe COVID-19 patients after ivermectin-based combination treatment: 55-62% normalization within 12-24 hours

Jacqueline C Stone College of Primary Care Physicians of Zimbabwe, Harare, Zimbabwe Pisirai Ndarukwa National Coordinator Ivermectin Research (MoHCC), Harare, Zimbabwe David E Scheim (  dscheim@alum.mit.edu ) US Public Health Service, Commissioned Corps, Inactive Reserve, Blacksburg, VA, US https://orcid.org/0000-0001-6841-7054 Barry M Dancis Independent bioinformatics consultant, Silver Spring, MD, US Jerry Dancis Department of Mathematics, University of Maryland, College Park, MD, US Martin G Gill ENT Surgeon, Private Practice, Fourways Life Hospital, Gauteng, South Africa Colleen Aldous College of Health Sciences, University of KwaZulu-Natal, Durban, South Africa https://orcid.org/0000-00027199-9160

pre-and post-treatment comparisons of SpO2 values. For this retrospective study conducted in Zimbabwe, ironically, challenges that constrained treatment capabilities also provided the opportunity to track changes in SpO2, all on room air, for 34 severe COVID-19 patients within 12 to 48 hours after beginning IVM treatment (see Figure 1). At the same time, these challenges tested whether severe COVID-19 patients with limited opportunities for hospitalization could be successfully treated at home or in clinics with rudimentary facilities.
Zimbabwe, a landlocked country in Southern Africa that shares a border with South Africa, had its rst reported case of COVID-19 in March 2020. 4 Eight cases and one death from COVID-19 followed in the same month. 5 The rst death occurred in the Wilkins hospital, Zimbabwe's main COVID-19 treatment center in the capital city of Harare. Facilities were limited at this hospital at the time; no capacity for ventilation was available. Given the rapid increase in COVID-19 patients in Harare after March 2020, 6 additional medical facilities began treating them.
COVID-19 wards were created at a general practice clinic by converting two staff rooms into a 4-bed ward and a storeroom into a 2-bed ward. Available equipment included several oxygen cylinders, an oxygen concentrator, six beds, and three monitors for SpO2 and blood ow parameters. The staff consisted of the lead author and another primary care physician who was off duty for several weeks after being injured in a vehicle accident on July 27, 2020, and either one or two nurses at different times, each on 12-hours shifts. During the initial months of the pandemic, in the absence of proven therapies and protocols, the standard of care evolved through early August 2020 to include corticosteroids, clopidogrel, aspirin, enoxaparin, rivaroxaban, a nebulized nano-silver preparation, zinc sulfate, hydroxychloroquine, azithromycin, doxycycline and in some cases an IV antibiotic.
However, the e cacy of these treatments was found to be limited, and by the end of July 2020, several COVID-19 deaths were recorded in the country. Based upon reports of initial success using IVM for COVID-19 treatment from colleagues in Johannesburg, South Africa, the College of Primary Care Physicians of Zimbabwe (CPCPZ) adopted and included IVM in their COVID-19 treatment protocol from August 8, 2020, starting initially with a 10-12 mg stat dose. Treatment of COVID-19 with IVM continued after the rst patients showed improved outcomes, with more rapid recoveries achieved at doses higher than the standard of 200 ug/kg as initially used.
In August 2020, after it became apparent that IVM added to standard of care was signi cantly reducing the death rate, together with the hospital system being overwhelmed, CPCPZ physicians decided to treat COVID-19 patients where an IVM-based protocol could be administered, including at local general practice clinics which had nursing care and oxygen, and at some patients' homes with nursing support and oxygen supplementation as available. As knowledge of this successful treatment regimen spread in Zimbabwe, other physicians began offering the same treatment, with improved outcomes, which led to the formation of the Zimbabwe COVID Front Line Clinicians Society.

IVM for COVID-19 treatment
The decision to include IVM in COVID-19 treatment protocols in Zimbabwe was made as the pandemic swept through that nation, overwhelming limited clinical care facilities, with no drug developed to treat COVID-19 being generally accepted as effective. A published case-controlled study of IVM treatment for COVID-19 conducted at four US hospitals 7 that had been initially released in a preprint in June 2020 found a 40% reduction in mortality among 173 patients treated with low dose IVM vs 107 case-matched controls (15% vs 25.2% deaths). Interest in IVM was supported by its Nobel prize-honored pedigree and its extensive use to treat a variety of human diseases in over 3.7 billion doses worldwide since 1987. 8-10 Another favorable characteristic of this drug is its extraordinary record of safety, well tolerated at high doses, 11,12 including in studies for COVID-19 treatment. 13,14 It is generally non-toxic even at doses far exceeding the therapeutic range. 15,16 Since August 2020, inpatient and outpatient treatments of COVID-19 with IVM have been applied across 25 countries, 10 with more than 20 RCTs conducted for IVM treatment regimens. 10,17,18 Seven of nine meta-analyses of these RCTs for IVM treatment reporting in 2021, all conducted using Cochrane analysis methodology, found signi cant 18-22 or possible 23,24 indications of IVM e cacy, with a mean 0.33 relative risk (RR) of mortality vs controls. Most of these 20 RCTs for IVM treatment of COVID-19 showed statistically signi cant mortality reductions or other clinical bene ts. Among the most recent and detailed of the nine meta-analyses noted above reported a pooled total 67% reduction in mortality for IVM vs controls, with a statistical signi cance for an overall effect of p=0.005. 20 A comprehensive review of the entire body of clinical studies for IVM treatment of COVID-19 by the Nobel co-laureate for IVM, Dr Satoshi Omura and colleagues, concluded that IVM yielded major reductions in mortality. 10 Two animal studies of IVM treatment at low human-equivalent doses, one for the SARS-CoV-2 virus in golden hamsters 25 and another for a related betacoronavirus (MHV-A59) in mice, 26 found statistically signi cant treatment bene ts, consistent with those found in the RCTs noted above. The indicated biological mechanism of IVM, competitive binding with SARS-CoV-2 spike protein, 27 is likely non-epitope speci c, possibly yielding full e cacy against emerging viral mutant strains.
The demonstrated safety of IVM at much higher than standard doses 11-14 allowed the latitude for dose escalation for IVM treatment of COVID-19 over time. On September 19, 2020, the CPCPZ held a seminar at which the use of a combination of IVM, doxycycline and zinc was presented, along with aggressive diabetes control, steroid use and anticoagulation, and this was suggested as the most effective and affordable care available at the time. Afterwards, combination therapy centered around IVM plus doxycycline and zinc became the standard COVID-19 treatment protocol used by the CPCPZ. The potential e cacy of these adjuncts was later supported by successful clinical trials results with treatments using IVM in combination with doxycycline 28 or with doxycycline and zinc. 29 This combination therapy for COVID-19 has been researched and advanced by Thomas Borody, 30 who in 1990 published the rst clinical trial of using a triple therapy of three inexpensive repurposed drugs for H. pylori, 31 the underlying bacterial cause of peptic ulcers. This triple therapy of repurposed drugs became the worldwide standard of care for peptic ulcers a decade later, after the patents for the palliative drugs Tagamet and Zantac expired, and the discovery of H. pylori as the cause of peptic ulcers was honored with the Nobel Prize for Medicine in 2005. 30

Methods
This study is a retrospective review of clinical data collected during the course of treatment of COVID-19 patients with therapeutic agents selected by their physicians to offer the greatest chances for clinical bene ts and recovery.

Outcomes
Outcomes tracked were 1) changes in SpO2 values from within one hour before treatment to 12, 24 and 48 hours after treatment, for a set of 34 COVID-19 patients for whom a pre-treatment SpO2 value and at least one SpO2 value up to +48 hours after start of treatment ( rst IVM dose) were available, all obtained on room air; and 2) mortality/deterioration: recovery, deterioration to critical status, or death for a larger set of 92 COVID-19 patients. For most of these patients, blood values for lymphocyte count, LDH, D-Dimer and CRP were also recorded, but complete blood test results were not obtained for every patient and therefore were not analyzed.

Participants
Sixty of the patients analyzed in this study, including all 34 in the SpO2 tracking series, were from Harare, treated by CPCPZ physicians either at local clinics or at patients' homes. Additional data for the mortality/deterioration series were obtained through inquiries sent on March 10, 2021, to all in a WhatsApp group of CPCPZ physicians treating COVID-19 with IVM asking for records of any of their patients so treated. Six physicians responded by furnishing records of 32 such patients, grouped with the 60 patients treated by CPCPZ physicians to comprise the mortality/deterioration series. For most of these six physicians, not every patient treated to recovery without incident was included, but it was con rmed that no other patient treated by them with IVM had died. Under pressures of patient care during the pandemic, record-keeping tended to be most comprehensive for the sickest patients, who thus are represented disproportionately in this study. For the mortality/deterioration analysis, the set of patients was further restricted to those with pre-treatment SpO2 of 90% or below (and a minimum SpO2 for all patients of 51%, as noted above). Table 1

Treatment
In the 92-patient mortality/deterioration series and in the overlapping 34-patient SpO2 tracking series, every patient received IVM at dosages described below in addition to selected other agents from the standard of care before August 2020. These other agents included corticosteroids, clopidogrel, aspirin, enoxaparin, a nebulized nano-silver preparation, rivaroxaban, zinc sulfate, azithromycin, doxycycline, and in some cases, an IV antibiotic. Patients treated in a clinic were assessed by a nurse upon admittance, with blood drawn and PCR tests conducted as feasible given the patient's condition and with severe symptoms necessitating immediate treatment. For those patients who contacted a CPCPZ physician from home requesting treatment, an online questionnaire was rst completed by the patient, after which, if COVID-19 was still a suspected diagnosis, a nurse visit to the home was conducted, with associated follow up per the procedures described for in-clinic patients.
As evidence of IVM safety and tolerability accrued following its use beginning in August 2020, its stat dose of 10 mg as used for the earliest patients was increased on September 11, 2020, to 10-12 mg every four days for three doses. Subsequently, the dosage was further increased to 12 mg IVM on the day of admission and then on days 4 and 8 plus doxycycline (100mg b.i.d.) and zinc sulfate (60mg/day). The latter regimen was used up through December 2020, when the second pandemic wave emerged in Zimbabwe. At that time, additional evidence of safety and tolerability of this regimen supported further dose escalation to a standard IVM dose regimen of 12 mg daily for ve consecutive days, with adjunct use of doxycycline and zinc sulfate continued at the doses noted. In some cases, for which this standard treatment regimen did not yield signi cant clinical gains within a few days, even higher doses of IVM were used, in some cases as high as 100 mg for a single dose. Transient adverse effects (AEs) such as blurred vision characteristic of high dose IVM often occurred at those dose levels, but no serious AEs associated with IVM were manifested in any patient. Each of the 34 patients in the SpO2 tracking series was treated with IVM, doxycycline and zinc.

Data collection
For patients in the SpO2 tracking series that were treated in clinics, values were tracked using monitors that continually displayed SpO2 values and readings for pulse rate and blood pressure and waveform images of blood pulses. For those treated at home, the intake nurse provided a pulse oximeter to the patient, unless the patient had one which the nurse deemed of reliable quality. The patient or a family member took SpO2 readings regularly, using the same oximeter as used for the pre-treatment reading. In most cases, these readings were taken daily, with much lower frequency after SpO2 values had risen signi cantly and the patient's clinical condition had correspondingly improved. Patients were instructed to message the nurse immediately if SpO2 ever decreased from a higher value to 93% or below or if any new clinical symptoms of concern developed. Patient outcomes were categorized as recovered, died, or deteriorated before recovery, as follows. Recovered: patients who recovered following IVM treatment and were still alive on September 1, 2021, three months or more after intake. Died: those who died following IVM treatment, whether or not the cause could have been unrelated to COVID-19. Deteriorated (before recovery): if at any time following the rst dose of IVM, any incident or condition, whether of indicated connection to COVID-19 or not, either caused them to be hospitalized or made that a reasonable course of action had that been an available and viable option.

Analytical methods
For the 34-patient SpO2 tracking series, a measure of percent normalization toward a fully optimal SpO2 value of 97 was applied, which for pre-and post-treatment SpO2 values S 0 and S 1 , is: 100*(S 1 -S 0 )/(97-S 0 ), capped at 100%. Pre-and post-treatment SpO2 values were plotted for all 34 patients using this percent normalization measure in Figure 1 and were also presented directly in Figure 2, Figure 3, and Table S2. Regarding the 92patient mortality/deterioration series, precise statistical comparisons of results from treatment vs control groups cannot be made other than in the context of an RCT. It is nevertheless of interest to compare the mortality results for this Zimbabwe mortality/deterioration patient series to the expected mortality of COVID-19 patients with similar characteristics. A review conducted in 2021 considered 46 prediction models for COVID-19 mortality, 33 two of which 34,35 were rated as having a low risk of bias; the same two were identi ed as being of the highest quality in another overview of COVID-19 mortality risk assessment models. 36 The risk assessment model used here is the one of these two that used pre-treatment SpO2 as one of its prediction variables. 35 37 This exceeds the overall CFR for the patient set of the 4C mortality risk predictor and indicates that using the 4C mortality risk predictor for the patient set in this study would not overestimate the risk factor.
The same UK consortium that developed the 4C mortality predictor also developed a model to assess the risk of clinical deterioration among inpatients with con rmed or highly suspected cases of  This deterioration model was developed using 73,948 patients recruited between February and August 2020, with clinical deterioration de ned as any requirement of ventilatory support or critical care, or death. This model was also applied to the Zimbabwe mortality/deterioration series, and predicted deterioration outcomes were compared with the number of patients who could be considered to have deteriorated, per the criteria speci ed above.
To calculate both mortality and deterioration probability estimates for the Zimbabwe mortality/deterioration series, three base variables were used: age, sex and pre-treatment SpO2 value on room air, while for mortality, the 4C model used the count of comorbidities as another variable, as was also recorded for the study patients. Abnormal values for respiratory rate, urea, CRP, lymphocyte count, and presence of radiographic chest in ltrates would have added extra points and would have increased associated probabilities to these risks calculated for mortality and/or deterioration. However, since only these four base variables were available for every patient, only these were used to calculate the 4C-predicted risks.
The comorbidities used in the 4C mortality predictor were the following: chronic cardiac disease, chronic respiratory disease (excluding asthma), chronic cardiac disease, chronic respiratory disease (excluding asthma), chronic renal disease, mild to severe liver disease, dementia, chronic neurological disease, connective tissue disease, diabetes, HIV or AIDS, malignancy, and clinician-de ned obesity. 38 For applying this risk predictor to the Zimbabwe mortality/deterioration patient series, these were the comorbidities used for the counts shown in Table S1. For some of these patients, certain variables used for the deterioration but not the mortality risk calculation would likely have signi cantly increased the deterioration probability values, and the calculated deterioration risk estimate is lower than that for mortality.
The 4C mortality and deterioration risk predictors were used to calculate probabilities for each of the 92 patients for mortality and deterioration, respectively. Although the Poisson distribution function can roughly approximate the probability of having a given number n or less total events from such a series having different probabilities, Monte Carlo simulation gives a much more precise probability estimate. 39 Monte Carlo simulations were executed, with 10,000,000 simulations performed ten times each for the mortality and deterioration estimations. These were performed using visual basic source code as listed in Supplementary File S3.

Ethics Approval
The Medical Research Council of Zimbabwe granted IRB approval (#E293) for this retrospective study.
Patients consented for their medical treatment and to having their de-identi ed data used in this study.  Table S2 show the progression of pre-treatment and post-treatment SpO2 values within 48 hours after rst dose of IVM for the 34 patients described above, all of whom were treated with IVM, doxycycline and zinc. All of these patients recovered. SpO2 values are shown at pre-treatment, all recorded within one hour before the start of treatment and at 12, 24 and 48 hours after treatment. The SpO2 value shown for a given patient at time x is that for the latest post-treatment time ≤ x. (Thus, for some patients, for example, having a post-treatment value within 12 hours before +24 hours but none in the next 24-hour period, the latest value at +48 hours is the same as that at +24 hours.) Figure 1 shows, for all 34 patients, SpO2 value changes from before IVM administration to post-treatment as percentages of full normalization to an optimal SpO2 value of 97 (95 is considered the minimum normal SpO2 value for a healthy child or adult by the US CDC 41 ). Red, orange and blue lines show, respectively, SpO2 values at +12, +24 and +48 hours post-treatment. The mean (±SD) SpO2 changes as this speci ed percent of optimal normalization were 55.1% ± 28.0% at +12 hours, 62.3% ± 26.3% at +24 hours and 64.3% ± 24.5% at +48 hours. As shown in Figure 1, these percentages of full normalization to SpO2=97 have a roughly uniform distribution across the full range of pre-treatment SpO2 values, from 66 to 93.  Table 2, those patients with lowest, mid-range and highest pre-treatment SpO2 values had mean SpO2 increases at +12 hours of 12.8, 5.4 and 2.8, respectively (all SpO2 values in percentage units). Figure 3 shows pre-and post-treatment SpO2 values throughout the entire observation period for each of the 34 patients. Note that these values were recorded less frequently after SpO2 had normalized.  Figure 4 shows successive SpO2 values for one patient who had a particularly rapid increase in these values after his rst dose of IVM. This patient was a 25-year-old male, treated by a CPCPZ physician at a GP clinic without a supplemental oxygen capability. He received his rst 12 mg IVM dose (repeated over the next four days) immediately after entering the clinic with respiratory distress and bilateral pneumonia indicated by stethoscopic examination. His COVID-19 diagnosis was con rmed by a positive result from a rapid antigen test. As shown, his SpO2 values increased from that recorded immediately before treatment (79%) to values at 45 minutes (87%), 90 minutes (92%) and 3 hours (95%) post-treatment. He was discharged later that same day, and his home SpO2 readings then uctuated between 92% and 95% over the next three days. By the fourth day after discharge, his SpO2 stabilized at 97%, his pulse dropped to 77 from prior values over 100, and he resumed working from home.

Results
For the 34 patients in the SpO2 tracking series, as shown in Figure 2, all but two had increases in SpO2 within the rst 48 hours after rst dose of IVM. As shown in Figure 3,

Results for mortality and deterioration
The mortality/deterioration series, as noted above, consisted of 92 COVID-19 patients with pre-treatment SpO2 values between 51% and 90%. Monte Carlo simulations were then executed, with 10,000,000 simulations run ten times. The number of simulated deaths in each run ranged from 7.079 to 7.081, with a mean of 7.080, and the mean of the associated individual standard deviations = 2.494. The probability of having zero to two simulated deaths in this series ranged in these ten runs was 0.0214 (identically for mean, minimum and maximum values).
Two patients other than the two who died t the deterioration criteria noted above. One was transferred to a hospital for three days because he became weak from not eating, and another was hospitalized for six days due to a drop in oxygen saturation. Both resumed IVM treatment after returning home and subsequently recovered. These two plus the two patient deaths yield a total of four deteriorations according to the 4C risk model for deterioration. 4C deterioration probabilities were calculated for each of the 92 patients, and Monte Carlo simulations were then executed for deterioration risk estimation, with 10,000,000 simulations run ten times. The number of simulated deteriorations in each run was 17.23 (identically for the mean, minimum and maximum of these values), with the mean of the associated individual standard deviations = 3.54. The probability of having zero to four simulated deteriorations in this series ranged in these ten runs from 0.0000196 to 0.0000216, with a mean of 0.0000207.
With the caveat above as to the limits of statistical analyses outside the context of an RCT, this calculation nevertheless suggests that the occurrence of four or fewer deteriorations among these 92 severe COVID-19 patients would be highly improbable under standard care. For comparison, note that in this series of 92 patients, 21 had pre-treatment SpO2 values ≤ 78%, and among these 21 patients, only two deteriorated (the patients who died). In sharp contrast were deterioration outcomes reported by Mukhtar

Discussion
This study is a retrospective review of clinical data collected amid the challenges of providing treatment with limited facilities and resources to COVID-19 patients with severe disease. Under such conditions, it was not possible to obtain blood test values for all patients, including values for lymphocyte count, LDH, D-Dimer and CRP, which were thus not analyzed. On the other hand, the lack of availability of oxygen supplementation for many of the patients treated resulted in the rare opportunity to track changes in SpO2 values all recorded on room air before and after administration of IVM to 34 patients, with several of these patients having presented with SpO2 values well below 90%.
The increase in SpO2 for these 34 patients as the percentage of fully optimal normalization to SpO2=97, as reported above, was (mean ±SD) 55.1% ± 28.0% at +12 hours, which rose to 62.3% ± 26.3% at +24 hours and then to 64.3% ± 24.5% at +48 hours after rst IVM dose. All but two of these 34  This nding offers clues as to the potential biological mechanism of IVM activity against SARS-CoV-2 since, for example, even an effective freeze on viral replication or rapid repair of damaged pulmonary alveoli would be unlikely to cause such rapid clinical improvements. One indicated biological mechanism of IVM activity, competitive binding with SARS-CoV-2 spike protein, as reviewed, 27 may, through a reversal of viral hemagglutination, act quickly to increase pulmonary capillary ow and in turn account for normalization of blood oxygenation. An additional plausible mechanism of IVM activity is the activation of the cholinergic anti-in ammatory pathway under the control of the vagus nerve, 51 which is regulated by acetylcholine and potentiated by the high-a nity binding of IVM (a positive allosteric modulator) to the alpha 7 cholinergic receptor a7nAChr 52 expressed on bronchial, vascular as well as to cytokine-producing cells (i.e., TNF, IL1 and IL6 secreting macrophages, lymphocytes and mast cells). 53 For the SpO2 tracking series, the pattern of rapid increases in SpO2 after start of IVM treatment as occurred for all but two of these 34 patients resulted in recoveries for all of them. This same pattern of highly successful outcomes extended to the 92-patient mortality/deterioration series, with recoveries of all but two patients. As noted above, although statistical signi cance cannot be determined through the 4C mortality and deterioration calculations, these odds calculations suggest that the probabilities for achieving the mortality and deterioration results obtained in this study were, respectively, low and extremely low under standard care.
No serious adverse effects (AEs) from IVM treatment were observed in any patient, although transient AEs such as blurred vision characteristic of higher-dose IVM administration were observed in some patients given doses as high as 100 mg. While comparative results using higher vs lower doses were not systematically tracked, the practice of increasing IVM doses for patients not initially responding to treatment worked out well and supports the indication that higher doses provide greater e cacy.

Conclusions
Pre-and post-treatment SpO2 values were recorded all on room air for 34 severe COVID-19 patients who were treated with the combination therapy of IVM, doxycycline and zinc plus other adjuncts. This application of multiple drugs against COVID-19 was based on Zimbabwe's experience with prior infectious diseases, for which early, aggressive use of multiple drugs has been a core treatment principle. For these 34 patients, all but two had increases in SpO2 from pre-to post-treatment, at every time interval of +12, +24 and +48 hours after rst IVM dose for which values were recorded. The mean increase in SpO2 value as a percentage of full normalization to SpO2=97 was 55.1% at +12 hours and 62.3% at +24 hours after rst IVM dose. These results closely parallel the mean SpO2 normalization of 65.2% as calculated from SpO2 values reported by Hazan et al. for 19 RT-qPCR con rmed COVID-19 patients having pre-treatment SpO2 ≤ 90, with all SpO2 values on room air.
The marked, rapid normalizations of blood oxygenation, p<0.0000001 for the 34-patient SpO2 tracking series in each time period analyzed (paired t-test), stand in sharp contrast to the well-established typical decline in SpO2 during at least the rst week after onset of moderate or severe COVID-19 symptoms and establish a cause-and-effect clinical bene t for IVM-based combination treatment of this disease. Furthermore, for the larger set of 92 severe COVID-19 patients treated with IVM and other adjunct agents, all having pre-treatment SpO2 values of 90 and below, all but two recovered, and only two of those recovering patients experienced deterioration before recovery. These two deaths and four deteriorations (including the two deaths) are much less than the expected seven deaths and 17 deteriorations predicted using the well-regarded 4C COVID-19 risk assessment model.
For the patients of this Zimbabwe study and the Hazan critical series, treatment at home or in clinics with basic facilities freed up hospital resources for other patients, and the treatment approach modeled in these studies could signi cantly relieve the pressure on overwhelmed health facilities. IVM is widely available worldwide, inexpensive, and one of the safest drugs in modern medicine, with its safety in "improving the health and wellbeing of millions" noted explicitly by the Nobel Committee in awarding its 2015 prize for the discovery of IVM. 54 These study results, therefore, support the extended deployment of IVM for COVID-19 treatment, complementary to immunizations for prevention.

Funding
No funding was received for this study. Pre-and post-treatment SpO2 values, all on room air, for 34 severe COVID-19 patients treated with IVM, doxycycline and zinc. Graphs A-I are in horizontal groupings by the range of pre-treatment SpO2 value and in vertical groupings by elapsed time after rst dose of IVM. (The lowest SpO2 value in the stipulated range of 51-93% was 66%.) For these elapsed times of x=12, 24, and 48 hours, the SpO2 value shown for a given patient is at the latest post-treatment time ≤ x. All pre-treatment SpO2 values are from within one hour before the start of treatment.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. JCStoneetal.2021.RapidincreaseofSpO2onroomairfor34severeCOVID19patients.Supplementarytables.pdf