EPIC-SR: The negative paxlovid data Pfizer has been sitting on

As mentioned in my most recent review, publication bias has been a major concern when trying to decide whether to prescribe paxlovid. Pfizer rushed to publish their positive study (EPIC-HR), but refused to release the results of a second simultaneous study (EPIC-SR) that was stopped (due to futility) at the exact same time. With the publication of EPIC-SR, we finally have that data, and as anticipated, it supports the current practice of not prescribing paxlovid (and therefore not testing for COVID) in the vast majority of patients. 

The paper

EPIC-SR: Hammond J, Fountaine RJ, Yunis C, Fleishaker D, Almas M, Bao W, Wisemandle W, Baniecki ML, Hendrick VM, Kalfov V, Simón-Campos JA, Pypstra R, Rusnak JM. Nirmatrelvir for Vaccinated or Unvaccinated Adult Outpatients with Covid-19. N Engl J Med. 2024 Apr 4;390(13):1186-1195. doi: 10.1056/NEJMoa2309003. PMID: 38598573 NCT05011513

The Methods

EPIC-SR is a phase 2-3 double-blind, randomized, placebo controlled trial. 

Patients

Adult patients presenting within 5 days of symptom onset with laboratory confirmed COVID-19.

They originally included patients with underlying conditions associated with an increased risk of severe disease if they had received a complete primary vaccination schedule, and unvaccinated patients if they had no underlying conditions. After receiving FDA approval for paxlovid, they changed their inclusion criteria, now excluding patients with any underlying conditions and focusing instead on patients who hadn’t received a vaccine in the last 12 months. 

Intervention

Paxlovid (300 mg of nirmatrelvir and 100 mg of ritonavir) every 12 hours for 5 days.

Comparison

Placebo. 

Outcome

Their primary outcome was time to sustained alleviation of symptoms. (In the final analysis they added “through day 28”. This change was made well after they stopped collecting data. It’s not clear why this change was made, and without access to the raw data, I don’t know if this change makes their numbers look better or worse, but given that everyone involved in this study is paid by Pfizer, you can make a pretty well educated guess.)

The Results

Despite reports 2 years ago that this trial was stopped early due to futility, they appear to have enrolled more patients than originally planned. They randomized 1440 patients, but according to their power calculation, only originally planned to enroll 1140. They then excluded 144 due to ‘data quality issues’, and lost a few more to follow up, and so report on 1250 patients in the final cohort. The cohort has a median age of 42, is about half female, and actually represents a pretty wide range of COVID presentations. About 20% already had severe COVID at the time of enrollment, and 50% had moderate COVID symptoms. As expected, the patients have fewer comorbidities than EPIC-HR, with 18% having a BMI over 30, 14% being smokers, 12% having hypertension, and 5% diabetes. Only 5% were over the age of 65. Honestly, though, considering how many young healthy patients I see with flu-like illnesses, this probably is representative of the patients I see, except only 56% were vaccinated against COVID. Despite being the “standard risk” trial, the authors state that 50% of this population was actually high risk. 

There was no difference in their primary outcome of sustained alleviation of symptoms, with patients taking 12 days to feel better in the treatment group and 13 days in the placebo group (p=0.60). They state the results were similar in both high risk and standard risk populations, but don’t present any data in the manuscript.

There was no statistical difference in the ‘key secondary outcome’ of Covid-19–related hospitalization or death from any cause (0.8% vs 1.6%, ARR -0.8%, 95% CI -2.0 to 0.4%). As discussed many times in these COVID trials, focusing only on “COVID-specific hospitalizations” is a fundamentally biased trial design that will favour the active treatment and downplay adverse events from the drug.

Despite widespread concern about ‘COVID rebound’, both viral rebound and symptomatic rebound were identical with active treatment and placebo.

Adverse events were statistically identical, but as is discussed below, the 24% adverse event rate in the placebo group highlights the problem with the way we try to identify and define adverse events in these kinds of trials. 

My thoughts

Industry runs good looking trials. This is double blind and placebo controlled. It is everything you could want, and it is negative. That is probably all you need to know for clinical decision making.

Of course, we know that although industry run trials look good on paper, they are all incredibly biased. When billions of dollars are at stake, you simply cannot leave the research in the hands of the people who will eventually shove that money into their Scrooge McDuck vaults. There is endless data that tells us that industry sponsored studies are biased in favour of their own product, and so despite using all the right EBM words, you must discount the conclusions of these financially conflicted papers. So knowing that, these negative results are even more damning. 

Does this trial prove paxlovid is useless? Of course not. No single trial can prove that, and there is still the possibility of benefit hidden in this data. (In relative terms, hospitalizations were 50% lower in the paxlovid group, and so the trial might simply be underpowered, or targeting the wrong population.) However, when assessed next to the other randomized data we have, and considering that the only positive data comes from the companies selling the drug, our clinical conclusion is easy: we shouldn’t be prescribing this drug, with very few exceptions.

What are the exceptions? It is really hard to know, because they mostly haven’t been studied, and so we have no way to know if paxlovid helps in any subgroup. I would still consider prescribing it in patients with significant immunosuppression, or very high risk comorbidities, like those being actively treated for cancer. For vaccinated patients who have significant co-morbidities, I discuss it, but in most cases, I still don’t think the evidence is strong enough to suggest it. The one thing I know for sure is that I will definitely not prescribe it to the patients included in EPIC-SR, including anyone who has had 2 doses of COVID vaccine, and anyone who lacks risk factors. 

(It is worth noting that although the EPIC-HR study was positive, the results were very underwhelming, and when factoring in the clear bias of the trial being run by the same company that stands to make billions on this chemical compound, was definitely not enough evidence on its own to warrant the use of paxlovid.)

There are a couple aspects of the methodology of this trial that might deserve deeper thought. They changed their inclusion criteria after the FDA approved paxlovid. In the methods section, they state that they did this for “clinical balance” with no further explanation. Scientifically, I can’t see a great reason for altering the study part way through just because the drug is being sold, so the cynic in me assumes this was a purely commercial decision in an attempt not to undermine sales of the drug. They slightly altered the definition of their primary outcome, without a great explanation. There are a number of secondary outcomes listed on clinicaltrials.gov that didn’t find their way into the manuscript (a form of publication bias).

The population included was mixed, and changed with time, and so it will be difficult to perfectly extrapolate these results. Perhaps the biggest question is whether waning immunity from vaccines matters. (I see a ton of patients who got 2 shots early on, but nothing since.) This study includes patients with any length of time from their last vaccine. Therefore, this population should include many patients with waning immunity, and probably represents the best case scenario for paxlovid. Similarly, almost 75% of these patients were enrolled with less than 3 days of symptoms, which means the results here should be better than we would expect in real world settings. 

As far as adverse events go, there was not an overall difference. (There were more adverse events thought to be due to paxlovid than placebo, led primarily by dysgeusia and diarrhea). However, as placebo has no biologic effects, an adverse event rate of 24% tells you something about the way that we define and record adverse events in these trials. A placebo has a 0% adverse event rate. (We can debate the nocebo effect if you want, but placebos have no measurable objective impact on the body.) These trials try to record every possible symptom as an adverse event, which seems like it would be a good thing, but it is actually a subtle way for the trial to downplay real adverse events caused by the active drug. If the active drug has a 2% rate of adverse events, those events will be completely lost in the noise of a quarter of all patients having adverse events, whereas the 2% would stand out against the 0% adverse event rate that placebo really has. Either way, RCTs, and especially those run by drug companies, always underestimate adverse events, so I wouldn’t use these specific numbers to counsel patients. 

There will be a group of physicians loudly advocating for paxlovid on Twitter, based primarily on the theory that although it fails to produce any measurable benefit in acute outcomes, it might be a savior when it comes to long-COVID symptoms. I appreciate that people are advocating on behalf of these long COVID patients. Just because hospitalizations and deaths are decreasing does not mean we should eliminate infection control measures that keep people safe from easily preventable infectious diseases. However, well meaning patient advocacy is not the same as science. Wishful thinking is not science. At this point, there is absolutely no credible evidence that paxlvoid has any effect on the long term consequences of COVID. If you want to advocate for this population, you should be advocating for an appropriately designed and funded RCT focused on long term outcomes. Until then, paxlovid should not be prescribed for long term outcomes because there is no good evidence of benefit, and all medications have harms. 

The biggest question is why the medical community allows the pharmaceutical industry to take advantage of us like this. Why do we let them test their own drugs? Why do we buy drugs from them after one unconvincing trial rife with financial conflict of interest? Why do we prescribe a medicine when we know that the company is sitting on negative data? It is easy to blame “big bad pharma” for these mistakes, but we know their goal, and it has nothing to do with helping patients. Their only goal is the maximization of profit. It is our job to help patients, and so if anyone is to blame for the billions of dollars wasted and millions of patients needlessly exposed to a unnecessary medication, it is us and our guideline committees that made treatment suggestions despite woefully inadequate data.

Bottom line

EPIC-SR is a double-blind placebo controlled RCT that shows no benefit from paxlovid in a vaccinated population, and in unvaccinated patients without risk factors. This will not change my practice, as we have known this trial was negative for 2 years (despite Pfizer trying to hide the data from us), and so I had already incorporated the results into practice. However, if you have been prescribing paxlovid in those populations, now is the time to stop. (For most of us, that also means you should no longer be doing any form of viral testing, as it provides absolutely no benefit to your patients.)

Other FOAMed

COVID therapy: There is still no evidence for nirmatrelvir/ritonavir (Paxlovid)

EPIC-HR: Some underwhelming data on Paxlovid

Paxlovid evidence: still very little reason to prescribe

Paxlovid does not reduce symptoms, definitive Pfizer trial finds.

Evidence based medicine is easy

The EBM bibliography

Evidence based medicine resources

EBM deep dives

References

Hammond J, Fountaine RJ, Yunis C, Fleishaker D, Almas M, Bao W, Wisemandle W, Baniecki ML, Hendrick VM, Kalfov V, Simón-Campos JA, Pypstra R, Rusnak JM. Nirmatrelvir for Vaccinated or Unvaccinated Adult Outpatients with Covid-19. N Engl J Med. 2024 Apr 4;390(13):1186-1195. doi: 10.1056/NEJMoa2309003. PMID: 38598573