Paxlovid evidence: still very little reason to prescribe

We are once again buried in a wave of viral respiratory illnesses, and much of the burden of illness is still COVID. I did a pretty thorough review of all COVID antiviral therapies in 2022, both on First10EM and on the EMCases Journal Jam, but that was 2 years ago. I think most of us are still confused about the exact role that paxlovid (nirmatrelvir–ritonavir) plays in the management of COVID, so I figured it was time to tackle that topic again. 

I reviewed EPIC-HR when it was originally published. (Hammond 2022) You can read the details in that post, but the quick summary is that this industry run, placebo controlled RCT of paxlovid in 2246 unvaccinated adult patients with less than 5 days of COVID symptoms was statistically positive, but uses a fundamentally biased composite outcome and an odd statistical analysis, among a number of other sources of bias. There is a hint of benefit here, but it is very hard to trust given the pharmaceutical industry’s history of lying to us through a veneer of science. More importantly, the results of EPIC-HR apply to very few of our patients these days (unvaccinated, but also mostly immune naive patients).

The elephant in the room is EPIC-SR. This study was run simultaneously to EPIC-HR, but was stopped early due to futility after enrolling 1440 patients. Despite stopping at the same time as EPIC-HR (which was published with tremendous fanfare), EPIC-SR has still not been published. This is very problematic, as the patients I am seeing everyday look more like the patients in EPIC-SR than those in EPIC-HR. There are some results posted on clinicaltrials.gov, and as expected (given that Pfizer refuses to publish them), they are not flattering. Days until recovery were identical between the two groups (12 vs 13), but adverse events severe enough to discontinue treatment were worse with paxlovid (2.4% vs 0.8%). There are a slew of other secondary outcomes, but I see no benefit. And again, this is an industry run study with about as much financial conflict of interest as is possible. (Look up the sales numbers for paxlovid, and you can understand why Pfizer wouldn’t want to publish negative data.)

There are a couple other studies of paxlovid. There is an unblinded RCT looking at 264 adult patients with severe comorbidities admitted to hospital with less than 5 days of COVID symptoms, comparing paxlovid to standard care. (Liu 2023) Their primary outcome was all cause mortality, and there was no difference (8 deaths in the control arm and 5 with paxlovid, p=0.32) There also were no differences in any of the secondary outcomes, such as hospital length of stay, ICU length of stay, or mechanical ventilation. This is an unblinded study, and it is too small to exclude a real difference, but it is the only study free of industry conflicts, and it is negative.

There is another small RCT that is published in Russian in a journal not registered on PubMed. (Balykova 2022) The study protocol was not registered prior to running the trial. They enrolled 264 patients with mild to moderate COVID and less than 5 days of symptoms, and randomized them to paxlovid or standard therapy. The trial was not blinded, and is also a drug company run trial. “Progression to a heavier severity level” occurred in 6% of the standard therapy group and 0% of the paxlovid group, so this was a positive study. So overall, we have an unblinded study, without a preregistered protocol, with financial conflict of interest, which demonstrates a small change in a poorly defined subjective outcome. This is an incredibly low level of evidence.

The Cochrane review on paxlovid currently concludes that there is “low-certainty” evidence that paxlovid reduces the risk of all‐cause mortality and hospital admission or death in high‐risk, unvaccinated COVID‐19 outpatients, and “very low-certainty” of benefit in inpatients. (Reis 2023) I am not sure there was a lot of value in the meta-analysis, as they are mostly just restating the results of EPIC-HR. Personally, I think those certainty levels need to be knocked down at least another level considering the significant financial conflict and the fact that we know more data exists and is being hidden. 

Overall, you could argue that the preponderance of the published data is positive, and therefore it might be reasonable to use paxlovid despite the significant uncertainty. However, the counter argument is that the published data is positive precisely because of publication bias, and we know that negative data is being hidden from us. Combined with the fact that the data we have is very weak and generally conflicted, there is a strong argument to be made that no one should be using paxlovid. I fall somewhere in the middle, but realistically much closer to the second argument than the first. I will offer paxlovid to high risk unvaccinated patients, but generally counsel almost all other patients against using it.

Other trials you might hear about

Cao (2023) compared VV116 to paxlovid in a non-inferiority RCT, and found it to be non-inferior. I am not sure that tells you much of anything if paxlovid itself might not even be better than placebo. The Cochrane review raises a concern about this paper, in that they started enrollment without ethics approval. 

There are now numerous observational studies. I am not sure it is worth going through each individual study, because the commentary will be rather repetitive. We don’t know why some patients were treated and others weren’t, which results in significant confounding, and makes the data incredibly uncertain. Overall, there doesn’t appear to be a dramatic effect. In one meta-analysis, the odds ratio for short term mortality was exactly 1, meaning paxlovid had exactly no detectable effect. (Tian 2023)

There were 2 retrospective publications (one inpatient and one outpatient) from the same authors in Hong Kong that looked at longer term outcomes and stand as outliers in demonstrating improved mortality. (Wong 2022a and Wong 2022b). Although these studies used propensity matching, you can only match for known recorded variables. These patients were all cared for in a public health system where paxlovid (or molnupiravir) was recommended, but only a tiny number (1% in one study and 10% in the other) were actually prescribed antivirals. What made these patients special? Might those differences be the real reason for the differences in outcomes seen? I think you can use this kind of observational data to expand or refine prescribing indications once you have a proven benefit in RCTs, but in the absence of RCT proven benefit, these studies are somewhat pointless. 

Conclusion

This data is awful. I don’t think we would have started using paxlovid if we weren’t in the middle of a pandemic. However, because of the pandemic, many people argued (including the Ontario Science Table) that we needed to accept a low level of evidence, and therefore paxlvod was recommended. Formally, these recommendations were temporary, intended to be readdressed when the pressures of the pandemic were less severe. Realistically, everyone on these panels knew that they would never have the time or energy to tackle these guidelines again, and so the acceptance of subpar evidence has become (predictably) permanent.

Like most topics in evidence based medicine, there is not a definitive answer. There are some hints of benefit, but those hints are tainted by bias and conflicts of interest.

Although EPIC-SR has never been published, we know it was negative (and can probably assume the results are worse than we know, given how drug companies operate). Therefore, one thing I think we can definitively say is that we should not be prescribing paxlovid to patients who would have qualified for EPIC-SR (unvaccinated adults without additional risk factors and high risk vaccinated people who have at least one risk factor for progressing to severe disease). 

Whether there is benefit in the higher risk populations remains to be seen, but is mostly irrelevant to my practice. Almost everyone I encounter with risk factors is vaccinated and almost all of the unvaccinated patients have no risk factors. For the rare unvaccinated patients with high risk comorbidities, I engage in shared decision making, which emphasizes the significant uncertainty in this data. 

Other FOAMed

Therapeutics Initiative: Paxlovid in British Columbia – Interim real-world analysis

EPIC-HR: Some underwhelming data on Paxlovid

Journal Jam 20: Outpatient Medications for COVID-19

The FIrst10EM COVID archives

References

Cao Z, Gao W, Bao H, Feng H, Mei S, Chen P, Gao Y, Cui Z, Zhang Q, Meng X, Gui H, Wang W, Jiang Y, Song Z, Shi Y, Sun J, Zhang Y, Xie Q, Xu Y, Ning G, Gao Y, Zhao R. VV116 versus Nirmatrelvir-Ritonavir for Oral Treatment of Covid-19. N Engl J Med. 2023 Feb 2;388(5):406-417. doi: 10.1056/NEJMoa2208822. Epub 2022 Dec 28. PMID: 36577095

Balykova LA, Selezneva NM, Gorshenina EI, Shepeleva OI, Kirichenko NV, Simakina EN, Kolontarev KB, Pushkar DY, Zemskov DN, Zaslavskaya KY, Noskov SM. Modern directed antiviral COVID-19 therapy: results of multicenter clinical effectiveness and safety study of fixed nirmatrelvir+ ritonavir combination. Pharmacy & Pharmacology. 2022;10(4):371-86.

Hammond J, Leister-Tebbe H, Gardner A, Abreu P, Bao W, Wisemandle W, Baniecki M, Hendrick VM, Damle B, Simón-Campos A, Pypstra R, Rusnak JM; EPIC-HR Investigators. Oral Nirmatrelvir for High-Risk, Nonhospitalized Adults with Covid-19. N Engl J Med. 2022 Apr 14;386(15):1397-1408. doi: 10.1056/NEJMoa2118542. Epub 2022 Feb 16. PMID: 35172054

Liu J, Pan X, Zhang S, Li M, Ma K, Fan C, Lv Y, Guan X, Yang Y, Ye X, Deng X, Wang Y, Qin L, Xia Z, Ge Z, Zhou Q, Zhang X, Ling Y, Qi T, Wen Z, Huang S, Zhang L, Wang T, Liu Y, Huang Y, Li W, Du H, Chen Y, Xu Y, Zhao Q, Zhao R, Annane D, Qu J, Chen D. Efficacy and safety of Paxlovid in severe adult patients with SARS-Cov-2 infection: a multicenter randomized controlled study. Lancet Reg Health West Pac. 2023 Apr;33:100694. doi: 10.1016/j.lanwpc.2023.100694. Epub 2023 Feb 6. PMID: 36777445

Reis S, Metzendorf MI, Kuehn R, Popp M, Gagyor I, Kranke P, Meybohm P, Skoetz N, Weibel S. Nirmatrelvir combined with ritonavir for preventing and treating COVID-19. Cochrane Database Syst Rev. 2023 Nov 30;11(11):CD015395. doi: 10.1002/14651858.CD015395.pub3. PMID: 38032024

Wong CKH, Au ICH, Lau KTK, Lau EHY, Cowling BJ, Leung GM. Real-world effectiveness of molnupiravir and nirmatrelvir plus ritonavir against mortality, hospitalisation, and in-hospital outcomes among community-dwelling, ambulatory patients with confirmed SARS-CoV-2 infection during the omicron wave in Hong Kong: an observational study. Lancet. 2022 Oct 8;400(10359):1213-1222. doi: 10.1016/S0140-6736(22)01586-0. PMID: 36216007

Wong CKH, Au ICH, Lau KTK, Lau EHY, Cowling BJ, Leung GM. Real-world effectiveness of early molnupiravir or nirmatrelvir-ritonavir in hospitalised patients with COVID-19 without supplemental oxygen requirement on admission during Hong Kong’s omicron BA.2 wave: a retrospective cohort study. Lancet Infect Dis. 2022 Dec;22(12):1681-1693. doi: 10.1016/S1473-3099(22)00507-2. Epub 2022 Aug 24. Erratum in: Lancet Infect Dis. 2023 Dec 5;: PMID: 36029795