Ivermectin for COVID-19 in adults in the community (PRINCIPLE): an open, randomised, controlled, adaptive platform trial of short- and longer-term outcomes

.hsva .va--anim-stagger :is(.va__row,.plot__row,.va__verdict){opacity:1}.hsva .va--anim .plot__shade{opacity:var(--shade-opacity)}.hsva .va--anim-stagger .ci{transform:translateY(-50%)}.hsva .va--anim-stagger .pt{transform:none}.hsva .va--anim-settle .pt{left:calc(min(var(--pt), 2) / 2 * 100%)}.hsva .va--anim-settle .ci{left:calc(var(--lo) / 2 * 100%);width:calc((var(--hi) - var(--lo)) / 2 * 100%)}

Ivermectin and the Research Cartel

Four Years Later – The Toughest COVID Pill to Swallow

Can Scientific Misconduct Be Criminally Prosecuted?

Significantly improved recovery and significantly lower risk of long COVID with ivermectin, despite very late treatment, low-risk patients, and poor administration.

36% lower ongoing persistent COVID-19 specific symptoms, p<0.0001 (details below). The primary recovery outcome shows superiority of ivermectin (probability of superiority > 0.999), missing from the abstract (details below). The p values for sustained recovery, early sustained recovery, alleviation of all symptoms, and sustained alleviation are all < 0.0001.

The efficacy seen here is despite testing about the worst possible usage: very late, low-dose, incorrectly administered, truncated treatment with low-risk patients. Recommended usage shows much better results.

The tables below compare the same principal investigator's trials of oseltamivir and molnupiravir, which show notable differences in design, analysis, and interpretation. A list of major issues in the trial follows.

c19early.org

Contrasting trial designs—ivermectin vs. molnupiravir

Trial design may favor detecting or not detecting efficacy. The contrast is clear in this case—these trials are for the same disease, from the same investigator, and overlap in time.

	Molnupiravir (PANORAMIC)1,2	Ivermectin (PRINCIPLE)3
Investigator	Prof. Chris Butler	Prof. Chris Butler
	Favors detecting efficacy	Favors not detecting efficacy
Treatment delay	up to 5 days from onset	up to 14 days from onset
Population	50+ or 18+ w/comorbidities — high-risk patients	18+ (mid-trial change) — low-risk patients
Treatment	5 days, 2x per day	3 days, 1x per day, dose below real-world use
Administration	Per recommendation (with or without food)	Opposite of recommendation for COVID-19 - without food, greatly reducing concentration4,5
Patients	183 patients per day (25,783)	11 patients per day (3,963 inc. concurrent control)
Publication delay	4 months	1.6 years (2.2 years from expected end)
Enrollment	Dec 2021 - Apr 2022	May 2021 - Jul 2022
Mutagenic	Yes6-20	No
Trial funder stance	Approved	Denied—efficacy confirms liability for prior errors
Cost	$70721	<$122
Merck profit	>$8B sales to date23, estimated $18 to produce24	No potential, many low-cost manufacturers

c19early.org

Contrasting trial designs—ivermectin vs. oseltamivir

Trial design, analysis, and interpretation may favor detecting or not detecting efficacy.

	Oseltamivir (ALIC⁴E)25	Ivermectin (PRINCIPLE)26
Investigator	Prof. Chris Butler	Prof. Chris Butler
	Favors detecting efficacy	Favors not detecting efficacy
Treatment delay	≤72 hours from symptom onset	≤14 days from symptom onset
Treatment	5 days, 2x/day	3 days, 1x/day, dose below real-world use
Administration	Per recommendation (with or without food)	Opposite of recommendation for COVID-19 - without food, greatly reducing concentration4,5
Primary recovery definition	Self-reported return to usual activities + fever/headache/muscle ache minor/absent; those minor symptoms and other symptoms of any severity may remain	Self-reported recovery, not limited to specific symptoms. Slower-resolving symptoms, e.g., those related to tissue repair rather than ongoing viral replication, may delay recovery and reduce sensitivity to improvement for other symptoms
Bayesian shrinkage	Priors shrink subgroup effects toward overall response; no strong null-centered overall treatment-effect shrinkage	Strong skeptical prior appears to shrink recovery HR toward 1; Prof. Sander Greenland estimates ~1.19 → 1.1527
Clinically meaningful threshold	Trial explicitly powered to detect a 1-2 day benefit; no additional meaningful threshold	HR ≥1.20 specified as meaningful for futility decisions
Recovery benefit	1.02 days faster [0.74-1.31]	2.06 days faster [1.00-3.06]
Long-term morbidity	Not tested	Significantly lower long-term morbidity
Interpretation	Recovered one day sooner; clinicians may consider treatment; accelerates recovery; even 0.7 days in a subgroup is described as an oseltamivir benefit	Unlikely to be clinically meaningful; further trials unwarranted

Severity	Issue (most recent update 30 days ago)
CRITICAL	1. 36% lower long COVID hidden in appendix
CRITICAL	2. Undisclosed null-biased Bayesian prior masks superiority
CRITICAL	3. Ad-hoc "clinically meaningful" threshold contradicted by same authors later in the same year for an ARI RCT
CRITICAL	4. False claims for long-term outcomes and recovery
CRITICAL	5. Significantly improved recovery is strongly associated with significantly lower mortality
CRITICAL	6. Superiority of ivermectin hidden
CRITICAL	7. Mid-trial allowance of mAbs/antivirals invalidates hospitalization/death data
CRITICAL	8. Superiority of budesonide not hidden
CRITICAL	9. Budesonide press release within 12 days
CRITICAL	10. Similar results, opposite conclusion
CRITICAL	11. Pre-specified "meaningful effect" only for interim futility
CRITICAL	12. Meaningful effect was 1.5 days for other arms
CRITICAL	13. Pre-specified "meaningful effect" probability changed 0.01->0.25
CRITICAL	14. Hospitalization/death probability of meaningful effect
CRITICAL	15. 10x lower accuracy in reported results
CRITICAL	16. Adverse event data missing
CRITICAL	17. Mortality results missing for concurrent control arm
CRITICAL	18. Details of hospitalization/death missing, many could be prior to treatment
CRITICAL	19. False claim on administration
CRITICAL	20. Results delayed 600 days
CRITICAL	21. Authors coverage of prior research extremely biased and cherry-picked
CRITICAL	22. Very late treatment
CRITICAL	23. Inclusion changed from 50+ to 18+ w/COVID dyspnea or comorbidity before start of ivermectin arm
CRITICAL	24. Mid-trial change to include lower risk patients
CRITICAL	25. Exteme conflict of interest
CRITICAL	26. Pause due to supply but medicine was stored at every study site
CRITICAL	27. Supply issue contradicated by manufacturer
CRITICAL	28. Design favors null result in contrast to molnupiravir trial by the same chief investigator
CRITICAL	29. Other arm results delayed over 5 years
CRITICAL	30. Inclusion changed from 7 to 14 days
CRITICAL	31. "Gate-keeping" protection of serious outcome evaluation
CRITICAL	32. Long delay between registration and enrollment
CRITICAL	33. Subject to participant fraud
CRITICAL	34. Inconsistent analysis - Bayesian vs. frequentist statistics
CRITICAL	35. Age 65+ reduced from 47% to 16% for ivermectin
CRITICAL	36. Contradictory interpretation of results (30 days ago)
SERIOUS	37. Fatigue not included in secondary outcome symptoms, dominates long COVID results
SERIOUS	38. Superiority and futility rules not mutually exclusive
SERIOUS	39. Delivery delay reduces perceived effect
SERIOUS	40. Recovery subgroup forest plot for all arms except ivermectin
SERIOUS	41. Even faster recovery with greater baseline severity
SERIOUS	42. Lack of recovery inverted to reduce effect size
SERIOUS	43. Slow delivery
SERIOUS	44. Administration on an empty stomach
SERIOUS	45. Mismatch with original proposal
SERIOUS	46. Eligibility criteria worse than concurrent favipiravir arm
SERIOUS	47. Recruitment questions varied
MAJOR	48. Ivermectin from source chosen has shown lower efficacy
MAJOR	49. Ability to pickup medication quickly removed from information sheet
MAJOR	50. Only three different doses, lower μg/kg dose for higher weights
MINOR	51. Duration == 7 missing
COMMENT	52. Efficacy not due to open label design

Responses: authors have not responded to any of these issues.

CRITICAL36% lower long COVID hidden in appendix. Page 358 in the appendix shows 36% lower ongoing persistent COVID-19 specific symptoms (p<0.0001) when combining the individual symptom results. The paper reports a 28% reduction (p=0.015), not mentioned in the abstract or conclusion. This appears to be a one of any symptom analysis, effectively increasing the weight of the more common “fatigue”, reducing the perceived effect (the difference does not appear to be due to adjustments - the adjustments in Table S6 to Table S39 make minimal difference). The effect may be diluted by including fatigue due to unrelated causes. If fatigue is excluded, the data on page 358 shows 41% lower long COVID, which may be more accurate. This is for very late and poorly administered treatment taken by only 89% of patients in a relatively low-risk population - benefits may be much greater with recommended usage and in high-risk patients.

CRITICALUndisclosed null-biased Bayesian prior masks superiority. Prof. Sander Greenland reports (starting on slide 38)27 that the Bayesian model applied an extremely tight, null-centered prior that was not reported in the paper. This prior pulls the frequentist hazard ratio for time-to-recovery down and slashes the posterior probability that the benefit reaches the trial's own “clinically meaningful” threshold of HR≥1.2. Even with the shrinkage, the posterior probability of any benefit exceeds 0.9999, well above the protocol's 0.99 superiority bar, however authors label ivermectin “unlikely to provide clinically meaningful improvement.” Greenland argues that the unexplained prior acts like a hidden penalty, lacks empirical justification, and biases the analysis toward the null, thereby allowing the authors to downplay results their own decision rules would otherwise classify as superior. Author notes the strong priors are not justified by previous trials and suggests that they reflect social pressure to discredit ivermectin.

CRITICALAd-hoc "clinically meaningful" threshold contradicted by same authors later in the same year for an ARI RCT. In the Immune Defence RCT on common acute respiratory illness (ARI)64, with two of the same senior investigators including the PI for PRINCIPLE, a similar reduction in illness-days from over-the-counter nasal sprays is highlighted as a clear therapeutic success without any extra “clinically meaningful” hurdle, is referred to as "clinically significant", and is promoted in the abstract and discussion65. By contrast, PRINCIPLE imposes a bespoke threshold and a very tight sceptical prior. This inverts the expected risk-benefit logic: improved recovery and minimizing viral replication is likely to be more significant for COVID-19, which showed higher risk of severity, systemic viral spread, and long-term sequelae.

CRITICALFalse claims for long-term outcomes and recovery. Authors claim that ivermectin is "unlikely to provide clinically meaningful improvement in recovery, hospital admissions, or longer-term outcomes", which is contradicted by their results. 36% (or the 28% from the author's calculation) lower long COVID is clearly clinical and very meaningful - it would represent an enormous global reduction in morbidity if adopted. The significantly faster recovery is also clearly clinically meaningful.

CRITICALSignificantly improved recovery is strongly associated with significantly lower mortality. Authors report highly statistically significant improved recovery but claim no clinical relevance. Across all 227 treatments we cover, improved recovery is very significantly associated with lower mortality, p<0.000000001 (from all studies that report both).

CRITICALSuperiority of ivermectin hidden. The protocol states "If the Bayesian posterior probability of superiority (a log hazards ratio greater than 0 corresponding to quicker recovery) for a treatment versus Usual Care is sufficiently large (e.g. ≥ 0.99), the null hypothesis will be rejected and the intervention will be deemed superior..". The intervention is superior (probability > 0.9999), yet there was no press release and immediate call for use, and this is not even mentioned in the abstract or conclusion.

CRITICALMid-trial allowance of mAbs/antivirals invalidates hospitalization/death data. Authors note: "From 16 Dec 2021, a minority of extremely clinically vulnerable patients could also access antiviral treatment or a monoclonal antibody infusion". However, there is no information on treatments provided or procedures for determining eligibility. This change invalidates hospitalization/death data after 16 Dec 2021. Hospitalization/death events occured in a small minority of patients and are expected to be strongly biased towards the extremely clinically vulnerable patients. Patients randomized to usual care are more likely to obtain alternative treatment. During the trial extension period sotrovimab was the most common treatment, with paxlovid and molnupiravir also being used66. Sotrovimab showed very high efficacy during this period67,68. It is normal to provide details of other treatments used in cases like this, the lack of disclosure suggests that the data confirms alternative treatment use significantly biased the results.

CRITICALSuperiority of budesonide not hidden. The same trial's budesonide arm did not hide the superiority: "There was a benefit in time to first self-reported recovery ... with a probability of superiority greater than 0·999, meeting the prespecified superiority threshold of 0·99" (in the abstract).

CRITICALBudesonide press release within 12 days. The superiority for recovery for budesonide was announced within 12 days of trial completion59,69. Slow but perhaps acceptable. Ethical and moral obligations mandate release as soon as possible. However for ivermectin, the results were hidden for around 600 days, and then misrepresented.

CRITICALSimilar results, opposite conclusion. Of the 6 arms reporting results (HCQ is still missing), three show superiority on the primary recovery outcome. Comparing ivermectin and budesonide, which both show probability of superiority >0.999 and similar results - several individual recovery results are better for ivermectin than budesonide - concurrent and COVID+: time to allevation of all symptoms, time to sustained alleviation of all symptoms, WHO5 wellbeing at day 28, all concurrent: time to allevation of all symptoms, time to sustained alleviation of all symptoms, and time to initial reduction of severity all show better results for ivermectin than budesonide. However the conclusions for each are the opposite - for budesonide authors concluded superiority, for ivermectin they conclude that ivermectin "is unlikely to provide clinically meaningful improvement in recovery". Note improvements are higher with ivermectin and the primary recovery outcome for several subgroups related to baseline severity - illness duration, baseline severity, and respiratory illness.

CRITICALPre-specified "meaningful effect" only for interim futility. The protocol (even the post-hoc versions) only mentions clinically meaningful effects in terms of interim futility analyses: "If the Bayesian posterior probability of a clinically meaningful treatment effect is sufficiently small (e.g. < 0.01) for the first co-primary endpoint (time to recovery), the intervention arm may be dropped from the study for futility". In the body of the paper authors do note that the pre-specified HR of ≥1.2 was only for futility evaluation, however the abstract drops this, implying that there was a pre-specified HR of 1.2 for superiority.

CRITICALMeaningful effect was 1.5 days for other arms. Authors did not mention a "meaningful effect" for budesonide, but added a "meaningful effect" of 1.5 days for colchicine, azithromycin, and doxycycline - which was sufficient to ensure very low probabilities for those arms. For ivermectin they show an improvement of 2.06 days, i.e., clinically meaningful according to the authors for all prior arms (for most other people, smaller improvements are also clinically meaningful, and as above the improvements translate into lower mortality for high-risk patients). Additionally, the poor design of the trial means the actual improvement for recommended usage is likely much greater.

CRITICALPre-specified "meaningful effect" probability changed 0.01->0.25. In June 2022, in view of interim (if not all at the time) results, authors changed the 0.01 probability to 0.25 (just above the 0.22 for the one outcome at the time). Authors note this change is only for ivermectin and favipiravir, and will return to 0.01 for future arms (appendix page 169). Authors claim a rationale for the change is in "Appendix A" but this appears to be missing.

CRITICALHospitalization/death probability of meaningful effect. For hospitalization/death authors previously used a threshold of 2% for calculating the probability of meaningful effect. For ivermectin they changed it to 20%. Consider the families of 20% or 2% of COVID-19 deaths, perhaps 4 million or 400,000 people based on an estimated 20 million total. Do they believe those deaths were not clinically meaningful?

CRITICAL10x lower accuracy in reported results. For budesonide authors reported key results with 10x greater accuracy, for example 10.9 vs. 13.3 days time to recovery, wherease for ivermectin all times are reported as integers, e.g., 14 vs. 16 days. This may be used to hide differences and to reduce efficacy (e.g., 3.6 vs. 5.4 becomes 4 vs. 5 and 3.6 vs. 4.4 becomes 4 vs. 4).

CRITICALAdverse event data missing. Other than a count of hospitalizations, no adverse event data was reported.

CRITICALMortality results missing for concurrent control arm. Authors do not provide the main mortality results for the concurrent control arm.

CRITICALDetails of hospitalization/death missing, many could be prior to treatment. Authors provide no details on the hospitalizations and deaths. Given the remote nature of the trial, the enrollment of very late stage patients, the very low event rate, and the very delayed time between patients signing up and actual enrollment indicated by some participants, many of the hospitalizations may have happened before medication was delivered and taken. This is supported by the subgroup analysis showing that patients >7 days from onset (likely >8-9 days to treatment initiation) contributed more to the ivermectin events. The patient information sheet for molnupiravir states that medication will be delivered by the next day2,70, while the patient information sheet for ivermectin has deleted "next day" only stating that medication will be delivered71.

CRITICALFalse claim on administration. Authors falsely claim that "the influence of food on absorption is not known"4. Guzzo et al.5 show that the plasma concentration of ivermectin is much higher when administered with food (geometric mean AUC 2.6 times higher). This is from 2002 and well-known among ivermectin researchers.

CRITICALResults delayed 600 days. Results were delayed around 600 days from the expected announcement time, with no reasonable excuse for hiding such positive results (or any results).

CRITICALAuthors coverage of prior research extremely biased and cherry-picked. Authors perform extreme cherry-picking on their discussion of previous research, and even then highly misrepresent those studies. For example, authors discuss the TOGETHER trial, without mentioning the known impossible data, refusal to release data despite pledging to, external sharing of results during the trial, randomization/blinding failure, and many protocol violations; and without mentioning that the principal investigator said that "There is a clear signal that IVM works in COVID patients.." in private72.

CRITICALVery late treatment. Patients were enrolled up to 14 days after the onset of symptoms. Extensive research for COVID-19 and other viral diseases show that early antiviral treatment is critical.

CRITICALInclusion changed from 50+ to 18+ w/COVID dyspnea or comorbidity before start of ivermectin arm. Inclusion was originally 50+ w/comorbidity or 65+, but was changed to 18+ w/COVID dyspnea or comorbidity or 65+ before the start of the ivermectin arm. The move to low-risk patients was specific to the ivermectin and favipiravir arms only73.

CRITICALMid-trial change to include lower risk patients. Inclusion criteria were modified mid-trial to allow enrolling anyone 18+, i.e. very low risk patients. This change is first seen in protocol 9.0 on July 12, 202144

CRITICALExteme conflict of interest. The chief investigator is also chief investigator for the PANORAMIC molnupiravir trial, with overlapping dates, and 7.2B+ financial conflict of interest between the two treatments.

CRITICALPause due to supply but medicine was stored at every study site. The trial claimed to pause due to a supply problem but medicine was stored at every site74, It is unlikely that all sites would have run out, if there was no supply in some locations, recruitment could have continued at other locations.

CRITICALSupply issue contradicated by manufacturer. The trial was paused with a reported supply issue, however the manufacturer stated that there were no supply issues.

CRITICALDesign favors null result in contrast to molnupiravir trial by the same chief investigator. Treatment delay, inclusion criteria, dosing, administration, and target size all show a design better for efficacy for molnupiravir, and worse for efficacy for ivermectin. Both trials have the same chief investigator and overlapping dates.

CRITICALOther arm results delayed over 5 years. The HCQ arm results were withheld for over 5 years55, consistent with extreme bias against reporting efficacy of low-cost treatments that were denied to the public by the funder.

CRITICALInclusion changed from 7 to 14 days. Inclusion was originally within 7 days of symptoms, but was changed to 14 days, compared to the molnupiravir trial which was started with 5 days75.

CRITICAL"Gate-keeping" protection of serious outcome evaluation. Authors declare a "gate-keeping" strategy to prevent evaluation of hospitalization/death if the recovery time difference is not significant4. Authors claim benefit for serious outcomes is unlikely without statistically significant benefit for recovery time, which is not logical, especially with low prevalence of progression - consider for example an intervention that prevented progression to mortality by 100%, but has no effect on resolution of a specific symptom, e.g., cough. The trial did not always have this gate-keeping strategy - protocol 4.0 had hospitalization/death as the primary outcome and protocol 5.0 added the new strategy (this was a post-hoc change for azithromycin and doxycycline related to the recruitment of low-risk patients and low event rates).

CRITICALLong delay between registration and enrollment. One participant reports filling out a form for the trial at the time of receiving a positive PCR result and not being called until much later on day 11 of COVID to complete enrollment76. A second participant reports waiting 9 days after online registration to receive an enrollment phone call77,78.

CRITICALSubject to participant fraud. There is no requirement for participants to have a face-to-face visit as part of trial participation. The self-reported design and the potential lack of professional medical examination results for many patients opens this kind of remote trial to participant fraud, which may be significant due to extreme politicization in the study country. Participant fraud has been reported for two other remote trials79,80, involving submission of fake surveys and repeated signups. Authors do not provide any information on attempts to limit participant fraud.

CRITICALInconsistent analysis - Bayesian vs. frequentist statistics. The protocol specifies Bayesian analysis which is used for some outcomes. However, authors have used frequentist statistics for other outcomes, with no known reason. This results in avoiding reporting Bayesian probability of superiority showing superiority of treatment for those outcomes.

CRITICALAge 65+ reduced from 47% to 16% for ivermectin. From the non-concurrent to concurrent populations (Table 1), we can see a dramatic change in the population. 47% of patients were over 65 in the control group prior to the ivermectin arm. For ivermectin this was reduced to 16% (focusing on low-risk patients is one method to reduce the chance of showing a benefit).

CRITICALContradictory interpretation of results. The same principal investigator ran an oseltamivir for influenza trial, which found 1.02 days faster recovery25. Authors suggest that clinicians may consider treatment, describe the result as accelerating recovery, and even describe 0.7 days in a subgroup as beneficial. In contrast, the 2.06 days benefit in this trial (and the 36% lower long COVID) is described as unlikely to be clinically meaningful, and authors state that further trials are unwarranted.

SERIOUSFatigue not included in secondary outcome symptoms, dominates long COVID results. The protocol (page 17 of 55, section 2.3.2.2, appendix page 119) does not include fatigue as a symptom. It's unclear if this was added later to dilute long COVID results (the included long-term follow-up SAP is post-hoc, dated June 2023). The more general fatigue, which may not be related to COVID-19, is by far the most common long-COVID symptom reported and dilutes efficacy—excluding fatigue ongoing persistent COVID-19 specific symptoms are 43% lower with treatment.

SERIOUSSuperiority and futility rules not mutually exclusive. The superiority and futility rules are not even mutually exclusive—the treatment can be considered superior and futile at the same time (superiority: Pr(HR > 1) ≥ 0.99, "meaningful-effect” futility: Pr(HR ≥ 1.2) ≤ 0.01"81).

SERIOUSDelivery delay reduces perceived effect. Recovery is defined as the time from randomization, however there are additional delays between randomization and delivery, and between delivery and the patient taking the medication. This has the effect of reducing the perceived effect of treatment. For example, the median time to alleviation of all symptoms was 4 and 5 days respectively, or 20% faster with treatment. If the delay until treatment is one day, this becomes 3 and 4 days, for 25% improvement, with progressively greater improvement with longer delays. The patient information sheet for molnupiravir states that medication will be delivered by the next day2,70, while the patient information sheet for ivermectin has deleted "next day" only stating that medication will be delivered71.

SERIOUSRecovery subgroup forest plot for all arms except ivermectin. The main paper shows a forest plot with subgroups of the primary recovery outcome for all arms (colchicine, budesonide, azithromycin, doxycycline) except ivermectin. For ivermectin, authors only show the hospitalization/death forest plot in the main paper. The recovery results show superiority of ivermectin.

SERIOUSEven faster recovery with greater baseline severity. Figure S2a shows that longer duration, at least one baseline major severity item, and respiratory illness all show greater improvement for recovery, consistent with the greater room for improvement in more severe cases. If authors truly believed that HR 1.2 is a required and valid threshold, they should not be ruling out use for higher-risk cases.

SERIOUSLack of recovery inverted to reduce effect size. Authors report the number of patients fully recovered at 3, 6, 12 months. It is typical to compare the risk of bad outcomes (failure to recover, hospitalization, death, etc.), however authors compare good outcomes. Authors method shows a significant improvement of 6% at 12 months, however the typical analysis shows a much larger 18% reduction in failure to recover. Consider a recovery rate of 90% in the control group, by the author's method it would be impossible for any intervention to create the HR ≥1.2 that they added.

SERIOUSSlow delivery. The patient information sheet for molnupiravir states that medication will be delivered by the next day2,70, while the patient information sheet for ivermectin has deleted "next day" only stating that medication will be delivered71.

SERIOUSAdministration on an empty stomach. Authors instructed patients that "no food should be taken two hours before or after administration"4. Guzzo et al.5, from 2002 and well known to ivermectin investigators, shows that the plasma concentration of ivermectin is much higher when administered with food (geometric mean AUC 2.6 times higher).

SERIOUSMismatch with original proposal. The original proposal for the trial starts with: "COVID-19 disproportionately affects people over 50 years old with comorbidities and those over 65 years old. The infection causes considerable morbidity and mortality in this population group in particular."82, yet authors later modified the trial to include anyone 18+.

SERIOUSEligibility criteria worse than concurrent favipiravir arm. In addition to inferior eligibility compared with molnupiravir, eligibility was even inferior to the concurrent favipiravir arm, with ivermectin further favoring a null result. As of July 8, 2021 favipiravir started at 50+ while ivermectin started at 18+ w/dyspnea or comorbidity as per the trial newsletter.

SERIOUSRecruitment questions varied. Recruitment questions varied, for example the video instructions for ambulatory care show the system asking about only two symptons - cough and fever. Notably, cough may be less responsive to treatment and increased enrollment based on cough may reduce the chance of showing efficacy84.

MAJORIvermectin from source chosen has shown lower efficacy. Authors chose to source ivermectin from Edenbridge, which ranked 7 out of 11 brands in In Vitro tests for antiparasitic efficacy85, requiring 5 days compared to 2 days for the best performing brand, and 3 days for 4 other brands.

MAJORAbility to pickup medication quickly removed from information sheet. Earlier versions of the patient information sheet (e.g., v3.186) allowed patients to pickup the medication from a local pharmacy instead of waiting for delivery. This was removed sometime before the ivermectin arm and the sheet now only lists delivery, excluding the possibility of very quick pickup of the medication after enrollment87.

MAJOROnly three different doses, lower μg/kg dose for higher weights. Only three different doses were used: 45-64kg (18mg), 65-84kg (24mg), and ≥84kg (30mg)4. Patients with higher weights will have progressively lower μg/kg dosing.

MINORDuration == 7 missing. The subgroup forest plot shows illness duration <7 and >7, without specifying what happens with == 7. The papers for other arms shows ≤7 and >7.

COMMENTEfficacy not due to open label design. Some people with strong prior claims of no efficacy have claimed that the efficacy here is due to the open label design. Moustgaard et al. showed there was no evidence to support this from analysis of 142 meta-analyses covering 1,153 trials. However, in this case we have multiple arms from the same exact trial that show this is not the case (which authors acknowledge in the paper, noting no evidence from the other arms). Moreover, any effect would be reversed because at the time and in the study country, the government and almost all media claimed that ivermectin was ineffective (and contrary information was censored).

The PANORAMIC trial for molnupiravir and the PRINCIPLE trial for ivermectin provide a good example of extreme bias in trial design. For molnupiravir investigators randomized 25,000 patients a median of 2 days from onset28. For ivermectin, they allow inclusion up to 14 days after onset — a delay incompatible with the recommended use of antiviral treatments, and incompatible with current real-world protocols. This delay alone would normally be more than enough to guarantee a null effect for an early treatment. However, authors also bias the population, treatment dose and duration, treatment administration, and sample size to favor a null result with ivermectin.

PANORAMIC and PRINCIPLE have the same chief investigator and primary contact1,3, and the molnupiravir and ivermectin arms overlap in time.

It is unclear why results were not released December 2021, why a reported supply issue was contradicted by the manufacturer, why the trial continued, and why results were delayed 19 months29.

Ivermectin was added to the PRINCIPLE trial on May 12, 202130 (June 2021 according to31), and favipiravir on April 26, 202130. 4,731 patients were enrolled as of April 8, 202132, by which time the azithromycin, doxycycline, and budesonide arms had completed. The colchicine arm had been running for one month and was later terminated with 156 patients. With an estimated enrollment of 1,000 per arm for ivermectin, favipiravir, and concurrent control, the trial would end when total enrollment reached around 8,000.

8,010 patients were enrolled and ivermectin was removed from the list of treatments under investigation on the website on or before Dec 2, 202133,34, suggesting that enrollment was complete and results would be available shortly thereafter.

By Dec 9 ivermectin was added back to the list35 with a note that the arm was paused due to supply issues. MedPage Today reported on the pause on Dec 1436. Notably, Merck's statement at the time shows a significantly softer stance compared to their previous comments36,37.

The reported supply issue is unusual - trials normally secure medication in advance, the reported trial manufacturer stated there were no supply issues38, investigators did not respond to journalist queries, there was reportedly no response to Freedom of Information requests39, alternate sources of ivermectin in the specified dosage were readily available, and there was no need for an identical match in appearance. The trial manufacturer was Edenbridge40, participants received standard standard foil strips41,42.

The trial later restarted the ivermectin arm. As of January 27, 2022, the trial was paused without explanation. As of February 11, 2022, the trial was open intermittently (twice daily between Sunday and Thursday), a change which further decreases the chance of participants receiving relatively early treatment. Delaying and restarting the trial at a later time may also reduce observed efficacy due to less severe variants in combination with the trial design.

We are pre-specifying subgroup analysis for enrollment up to Dec 2, 2021, for treatment within 2 days of onset, and for treatment of high risk patients (as originally defined by the trial).

Date	Change
PRINCIPLE trial timeline
March 22, 2020	Inclusion ≤7 days, age 50+ w/comorbidity or 65+.
June 16, 2020	Inclusion changed to ≤14 days.
February 14, 2021	Inclusion changed to 18+ w/COVID dyspnea or comorbidity or 65+43.
May 12, 2021	Ivermectin listed as current intervention in protocol30.
June 2021	Ivermectin added according to web site31.
Jul 12, 2021	Inclusion changed to 18+44.
December 2021	Anticipated completion of ivermectin arm.
December 3, 2021	Ivermectin arm ends, removed from web site between Dec 2 and Dec 334,45,46.
December 2021	No press release or rapid top-tier publication, indicating positive results.
December 9, 2021	Ivermectin added back to web site with claim of pause due to supply issues.
December 14, 2021	Trial does not respond to MedPage Today regarding supply problems. A statement from Merck is dramatically different to their previous position and is consistent with them knowing that a trial they cannot ignore has positive results and them being unsure if they can suppress the results47.
December 25, 2021	The trial supplier, Edenbridge, denies any supply issue. Prof. Chris Butler declines to comment48. The trial used standard widely available tablets41,42.
January 14, 2022	Prof. Paul Little, TSC chair, is removed from the trial in protocol version 1349. The TSC is responsible for reporting ethical issues.
January 27, 2022	Trial paused without explanation50.
February 11, 2022	Trial only open intermittently (twice daily between Sunday and Thursday), adding further enrollment delays51.
July 8, 2022	Extended ivermectin arm ends52.
July 2022	No press release or rapid top-tier publication, indicating positive results.
June 2, 2023	Sometime between May 2 and June 2, authors add a note on the web site indicating that, against protocol, they are delaying and will release in a rigorous and transparent way after extended 1 year followup ends in July53,54. Note that the analysis code for professional trials is written and tested in advance.
November 6, 2023	Links to the protocol, amendments, and other supporting documents were removed from the web site.
December 2023	Still no results or update. A link was added to a version 14 of the protocol dated August 8, 2022 (after all arms had completed). The link does not work, pointing to an internal University of Oxford site. The latest version available is 13.0, dated January 14, 202249 (during the ivermectin arm, after the expected end in December 2021).

c19early.org

PRINCIPLE trial selective reporting delays

Results for effective but politically inconvenient treatments were withheld for very long periods.

	Treatment patients	Duration	Results delay
HCQ55	413	2 months	Over 5 years55
Azithromycin56	540	6 months	56 days57
Budesonide58	1,073	4 months	12 days59
Doxycycline60	780	5 months	42 days57
Colchicine61	156	3 months	120 days62
Ivermectin26	2,157	14 months	600 days (810 days from ~1,000 per arm enrollment)
Favipiravir63	~2,250	15 months	780 days (1,000 days from ~1,000 per arm enrollment)

This is the 48th of 54 COVID-19 RCTs for ivermectin, which collectively show efficacy with p=0.000000027.

This is the 100th of 106 COVID-19 controlled studies for ivermectin, which collectively show efficacy with p<0.0000000001.

Standard of Care (SOC) for COVID-19 in the study country, the United Kingdom, is very poor with very low average efficacy for approved treatments89. The United Kingdom focused on expensive high-profit treatments, approving only one low-cost early treatment, which required a prescription and had limited adoption. The high-cost prescription treatment strategy reduces the probability of early treatment due to access and cost barriers, and eliminates complementary and synergistic benefits seen with many low-cost treatments.

Important missing comments or errors? Let us know.

risk of death/hospitalization, 1.0% higher, HR 1.01, p = 0.97, treatment 34 of 2,157 (1.6%), control 27 of 1,806 (1.5%), concurrent and eligible, primary outcome, details missing, mAb/AV confounding.

risk of death, 62.3% lower, RR 0.38, p = 0.18, treatment 3 of 2,157 (0.1%), control 12 of 3,256 (0.4%), NNT 436, non-concurrent, authors do not provide details for concurrent deaths.

risk of mechanical ventilation, 151.4% higher, RR 2.51, p = 0.63, treatment 3 of 2,149 (0.1%), control 1 of 1,801 (0.1%).

risk of ICU admission, 319.0% higher, RR 4.19, p = 0.23, treatment 5 of 2,149 (0.2%), control 1 of 1,801 (0.1%).

time to sustained recovery, 16.0% lower, HR 0.84, p < 0.001, treatment 2,157, control 1,806, inverted to make HR<1 favor treatment, concurrent and eligible.

early sustained recovery, 23.1% lower, HR 0.77, p < 0.001, treatment 2,154, control 1,805, inverted to make HR<1 favor treatment, concurrent and eligible.

sustained alleviation, 17.4% lower, HR 0.83, p < 0.001, treatment 1,826, control 1,535, inverted to make HR<1 favor treatment, concurrent and eligible.

alleviation of all symptoms, 16.0% lower, HR 0.84, p < 0.001, treatment 2,154, control 1,805, inverted to make HR<1 favor treatment, concurrent and eligible.

first reported recovery, 13.0% lower, HR 0.87, p < 0.001, treatment 2,157, control 1,806, inverted to make HR<1 favor treatment, concurrent and eligible, primary outcome.

no recovery at 3/6/12 months, 28.0% lower, HR 0.72, p = 0.02, treatment 94 of 1,941 (4.8%), control 109 of 1,624 (6.7%), NNT 54.

risk of no recovery, 17.6% lower, RR 0.82, p = 0.001, treatment 417 of 1,848 (22.6%), control 420 of 1,533 (27.4%), NNT 21, day 365.

risk of long COVID, 36.3% lower, RR 0.64, p < 0.001, treatment 1,886, control 1,567, all symptoms combined.

risk of long COVID, 63.2% higher, RR 1.63, p = 1.00, treatment 2 of 1,507 (0.1%), control 1 of 1,230 (0.1%), ongoing/persistent, fever.

risk of long COVID, 72.7% lower, RR 0.27, p = 0.33, treatment 1 of 1,819 (0.1%), control 3 of 1,489 (0.2%), NNT 683, ongoing/persistent, cough.

risk of long COVID, 50.1% lower, RR 0.50, p = 0.04, treatment 15 of 1,886 (0.8%), control 25 of 1,567 (1.6%), NNT 125, ongoing/persistent, dyspnea.

risk of long COVID, 35.3% higher, RR 1.35, p = 0.74, treatment 5 of 1,808 (0.3%), control 3 of 1,468 (0.2%), ongoing/persistent, chest pain.

risk of long COVID, 25.2% lower, RR 0.75, p = 0.36, treatment 21 of 1,831 (1.1%), control 23 of 1,501 (1.5%), NNT 259, ongoing/persistent, smell.

risk of long COVID, 58.7% lower, RR 0.41, p = 0.42, treatment 2 of 1,821 (0.1%), control 4 of 1,503 (0.3%), NNT 640, ongoing/persistent, diarrhoea.

risk of long COVID, 70.2% lower, RR 0.30, p < 0.001, treatment 10 of 1,739 (0.6%), control 27 of 1,400 (1.9%), NNT 74, ongoing/persistent, headache.

risk of long COVID, 29.8% lower, RR 0.70, p = 0.25, treatment 23 of 1,739 (1.3%), control 27 of 1,433 (1.9%), NNT 178, ongoing/persistent, muscle ache.

risk of long COVID, 47.1% lower, RR 0.53, p = 0.03, treatment 19 of 1,724 (1.1%), control 30 of 1,441 (2.1%), NNT 102, ongoing/persistent, generally unwell.

risk of long COVID, 20.1% lower, RR 0.80, p = 0.19, treatment 66 of 1,876 (3.5%), control 69 of 1,567 (4.4%), NNT 113, ongoing/persistent, fatigue.

risk of long COVID, 42.9% lower, RR 0.57, p < 0.001, treatment 1,886, control 1,567, all symptoms combined.