That’s not technically correct. It reduces the need for urgent care visit, ER visit or hospitalization by 90%. The majority of the patients were low risk and they don’t report the number actually hospitalized (half were under 44, so I doubt any needed hospitalization). Interestingly, the median time to symptom resolution was 3 vs 4 days; and almost all patients had resolution of symptoms within 2 weeks. These were not statistically significant. Levels of the virus were no different in the arms nor were episodes of hypoxia. The only endpoints that were different were subjective measures of illness and seeking medical care. The endpoints’ validity are compromised by the trial design: patients were not blinded to their treatment arm. As a result, the primary outcome being driven by patients’ opinion of their illness rather than specific physiologic measures makes the study difficult to interpret. That, coupled with the trial being under-powered with only 11 events between the two arms, makes it difficult to conclude anything other than use of an inhaler may reduce patients’ subjective need to seek medical care after developing mild COVID-19.
I am a bit surprised Oxford’s IRB approved the trial that was so sloppily designed. I think they went with a composite endpoint due to limited sample size. A better endpoint would’ve been hospitalization alone, or persistent hypoxia (main reason people are hospitalized). I think it’s a good idea to investigate given dexametbasone’s success in a much larger and better designed trial for severe covid, and there’s pressure to identify any signal of benefit, but in normal circumstances this study would be heavily flagged during peer review for the above flaws.
Aren't you a general surgeon? I'd imagine trial design is not your area of expertise. That being said the 1 day recovery difference WAS statistically significant (p = 0.007) and it was 7 vs 8 days, not 3 vs 4. The proportion of patients reporting symptoms at day 14 was also statistically significant 10 vs 30% (p = 0.003). Fever is also not a subjective report and was statistically different. You speak poorly of patient's ability to report symptoms, almost as if their perceptions have nothing to do with reality. I would assume that the daily phone calls / logs asked specific symptom related questions to gather the information. Not "You good bro?" Additionally, to what purpose would finding statistically significant difference in viral load matter if it still results in an improvement of a patient's course of illness?
Also, comments from the scientific community seem to favor the trial design as it is.
https://www.sciencemediacentre.org/expert-reaction-to-preprint-on-an-rct-of-inhaled-budesonide-in-the-treatment-of-early-covid-19/
Overall improvement at 14 days was not statistically significant, but if it decreases hospital visits it has a valid use. And it did list a "significant reduction of symptoms at 28 days" (long covid) in their discussion, despite me not being able to find specifics on it.
Is it a perfect study? No. Is it as bad as your quick response to it was? No also. Should further data be provided and further studies be done to confirm? Yes. Does it remain promising? Also yes.
I'm not a general surgeon but won't belabor that point and somewhat surprised by your anger. I have other qualifications related to my clinical practice. For the record, most general surgeons are quite familiar with trial design since surgical trials are quite difficult to perform correctly, so they are discussed at length before being implemented. That being said, are you a clinical trial specialist? If not, then we're in the same boat and entitled to opinions from the study. That is the point of peer review and examining all the data rather than headlines.
- I agree that median time to self-reported recovery was 7 vs 8 days, which was significant. You are right about that. Peak fever temperature was statistically different, but unclear what significance a mean 0.5 degree difference at one time point is from clinical or patient perspective. The overall fever curve of the patients (in the supplemental data) was not significantly different, though use of anti-pyretics was less in the budesonide group.
The Flupro data (a validated questionnaire used for influenza, so probably useful for Covid and more standardized) reported was 3 vs 4 days median recovery, % symptom resolution at 14 days and change from day 0 to 14. The latter two were not significant but the change was. However, the separation occurred after most patients had stopped therapy. Hard to know what to make of all that, since a validated questionnaire should be considered as better evidence than unvalidated questionnaires. That two ways of measuring similar outcomes showed differing results should be considered in interpreting the data. The overall finding is that some objective and semi-quantitative findings were significant, and others weren't. It muddies the waters, unfortunately. The other point about this is the early recovery suggests that the overall symptom course in these patients was mild, and again suggests that hospitalization was uncommon. I would think if that endpoint alone was significant, they would note it in the paper.
- I do agree that the patient's perceptions are important, but you cannot ignore that a subjective outcome will be affected by a non-blinded trial. That is the purpose of blinding in a randomized trial. I don't need to be an expert in trial design to know that, and can assure you from experience that it would be not ignored in peer review.
- The primary endpoint remains "COVID-19-related Urgent Care visit, Emergency Department assessment or hospitalisation". They didn't provide a breakdown (unless someone can find the supplemental data, which probably has it) so we can't say it was hospitalizations that were reduced. That is my other point and where the initial reporting is not technically correct. I will concede that Urgent care and ER visit reduction is a good thing, but again there are a lot of subjective factors that go into that outcome that can be biased by a non-blinding design.
Regarding the commentary from your link:
“There is very high qPCR confirmed diagnosis. However, the study is open without double-blind which could include bias in patient selection by using an age range that is younger and with limited co morbidities associated with increased risk of hospitalisation and death. However, it establishes a good principle, recovery being fast and limited adverse effects. Also the treatment could be economic and used worldwide – roll on a bigger study and broader age range and in populations with risk factors of severe COVID-19 pneumonia...Next step, a randomised and double-blind clinical study should be undertaken to confirm the findings and particularly including older age adults with more significant comorbidities, with the aim of starting treatment within 4 days of first symptom onset."
Further on: “The measurements and the primary outcome measurements look most effective to demonstrate early recovery – simply showing not requirement for secondary medical care (with defined activities). Concerns are the lack of double blinding the therapy which could involve bias, the inclusion of young adults at low risk of complicated disease and the limited number of individuals with comorbid conditions increasing risk of disease."
- I believe Dr Higenbottam was more eloquent at explaining my main concerns, and I presume he meets your criteria for expertise?
@OtherJen and
@former player : I agree with your points, the trial is promising but needs follow-up with a better design. I wish they had done a trial similar to the dexamethasone studies, which were well-powered and blinded. If they ultimately do, then it's worth it (inaccurate press releases aside). It is important that the university's press releases are as accurate as possible, to avoid the problems from early adoption using weak data last year. Already we see this trial being equated to reducing hospitalizations when in reality it seemed to reduce mild COVID symptoms in a population that does not usually require hospitalization.
ETA: Found the supplemental tables, they don't report what the breakdown of hospitalization vs. urgent care visits was.