Famotidine: Another Proposed Weapon Against COVID-19

My brother recently made me aware of a randomized trial testing whether famotidine (brand name: Pepcid) can prevent bad outcomes in patients with COVID-19.  The idea for this came from an infectious disease physician, Michael Callahan, who was working in China when the coronavirus outbreak hit and went to Wuhan to study (?) it.  Dr. Callahan and colleagues observed that “elderly peasants” were more likely to survive than their wealthier compatriots.  Looking further, they noticed that many “elderly peasant” survivors had chronic heartburn and were taking famotidine, while the wealthier Chinese citizens who had chronic heartburn were taking omeprazole, which works by a different mechanism (proton-pump inhibitor).  The patients on omeprazole died at a higher rate (although the difference between the famotidine and omeprazole users was not statistically significant).

When he came back to the United States, Dr. Callahan contacted some scientists he knew (details here) eventually involving a computational chemist who did some modeling which suggested famotidine might be effective against SARS-CoV-2.  He also worked with Dr. Kevin Tracey from Northwell Health near New York City to develop a randomized, double-blind trial to evaluate famotidine.  This trial is currently underway, with plans to enroll 1,174 patients and an interim analysis when results are available for the first 391 participants.

That’s a pretty big trial, and I wonder what the primary outcome(s) is and what magnitude of difference they are looking for that requires that sample size.  I can’t find the trial on clinicaltrials.gov, where almost all U.S. trials are registered as part of the requirements for publication in a peer-reviewed journal.  In order to have a decent chance at detecting the difference in mortality seen with the Chinese patients (27% vs. 14%), they would need roughly 155 patients in each group, or 310 total.

In the meantime, a paper examining a retrospective cohort of patients with COVID-19 and admitted to  Northwell Health has been uploaded to medrxiv.org (here). Some of the authors of the paper are involved with the randomized trial, including Drs. Kevin Tracey and Michael Callahan. 

Famotidine is a histamine H2-receptor blocker and its primary function with respect to the gastrointestinal tract is to inhibit gastric secretion.  With oral doses the intragastric pH increases from about 1.5-3.5 to 5.0-6.4, depending on the dose taken.  Its elimination half-life is affected by kidney function and the half-life of famotidine will increase in patients with kidney disease.  Adverse effects in the central nervous system can occur in patients with moderate to severe kidney impairment if doses are not reduced or spaced out at longer intervals. (See here for details).

Why would famotidine be effective against COVID-19?  According to the Sciencemag article, famotidine may bind to papain-like protease, which is important for viral replication.  According to the retrospective cohort paper, famotidine might bind with another protease, 3C-like protease, which appears to be another molecule according to this source. 3C-like protease is also involved in coronavirus replication.  I’m not a chemist so I can’t say if these two molecules can essentially be considered equivalent?  The source just mentioned states that papain-like protease has additional key functions that 3C-like protease does not have.

Why would famotidine be prescribed to patients in the hospital?  Through a quick literature search I found that physicians frequently prescribe proton-pump inhibitors like omeprazole and to a lesser extent famotidine to critically ill patients for “stress ulcer prophylaxis.”  Patients in the intensive care unit (ICU) undergo certain procedures, such as mechanical ventilation, or may have risk factors such as trauma or coagulopathy, which puts them at higher risk for bleeding in the upper gastrointestinal tract and subsequent bad outcome. However, in recent years critical care methods have improved and stress ulcers are relatively uncommon (4% or less), so there is some controversy about prescribing them.

Now to the retrospective cohort paper.  Why would the authors put out this paper when they predicted on April 27 that the results of the randomized trial would be known “in a few weeks.”  They appear to include different patient populations.  The retrospective paper’s inclusion criteria were adults with a positive polymerase chain reaction test for SARS-CoV-2 RNA with the test result available within 72-hours.  Patients were excluded if they died or needed “urgent or semi-urgent” intubation (mechanical ventilation, i.e., help breathing) within 48 hours of admission.  According to Sciencemag, patients in the randomized trial are “critically ill, many on ventilators.”

It is almost always problematic when the inclusion or exclusion criteria depends on a study outcome.  If for no other reason, you can never duplicate this patient population in a prospective manner since you can’t perfectly predict which patients who are sick enough to be hospitalized will not need mechanical ventilation within 48 hours.  It also can bias a study in favor of one of the treatment groups (e.g., “patient received an entire treatment course”) although it’s not obvious that’s the case here. So by study criteria the patients in this paper have milder disease than what can be predicted prospectively. 

Patients in the famotidine group had to receive it within 24 hours of admission, otherwise it was classified as “absent.”  Does this mean if the patient received it >24 hours after admission, they were in the non-exposed group?  The paper is silent on this matter, but if interpreted literally the >24 famotidine receivers were part of the non-exposed group.  The rationale for the 24-hour criterion is not given.  Also, since these patients were not critically ill at the time they received famotidine, it does not appear to have been prescribed for stress ulcer prophylaxis, unless the prescribing physician wanted to get them started on it “just in case”?  Only 15% of the famotidine, and 1% of the non-exposed patients had been on famotidine prior to hospitalization, so continuing home medication does not explain the in-hospital famotidine use either.

“The primary outcome was a composite of death or endotracheal intubation within 30 days of hospital admission.”  The authors collected data on other factors that might affect risk of intubation or death, such as age, body mass index, and co-morbidities like diabetes or hypertension.  They also captured whether the patient was given hydroxychloroquine or a proton-pump inhibitor like omeprazole.  Plasma ferritin levels were assessed (timing?) as a marker of “cytokine storm” (disproportionately strong reaction of the immune system to infection, often resulting in tissue damage and sometimes death).  They analyzed this data using some standard statistical methods appropriate for the study design (Cox proportional hazards).

The study included 1,536 patients “unexposed” to famotidine and 84 patients who received it within 24 hours of admission.  Although there were a pretty small number of famotidine patients, by characteristics measured in the study they were fairly similar to the unexposed patients (Table 1). Regarding Table 1, an error occurred where the number of patients receiving “Non-rebreather or similar” (last line in table) was transposed between the famotidine and no famotidine (unexposed) groups.  That is, 480 (31%) of no famotidine patients were on a non-rebreather while 21 (25%) of famotidine patients were on it.  The study biostatistician did propensity scoring to create better-matched study groups, although this was probably not necessary (but also not wrong).  The propensity scoring eliminated a lot of the no famotidine patients but the conclusions of the study didn’t change.

The results:  only 8 of the 84 famotidine patients died or were intubated (10%) vs. 332 (22%) of the no famotidine patients.  This was statistically significant when the Cox proportional hazards model was done.  The Cox model looked at the time to either intubation or death.  Kaplan-Meier curves (Figures 1 and 2) also show the famotidine patients doing better than the no famotidine patients.  There seems to be a discrepancy between the description of Figure 1 in the article text “composite outcome of death or intubation” and the title of Figure 1 “intubation-free survival.”  It seems to be intubation-free survival because there are no outcomes after 20 days in Figure 1 but there are in Figure 2 (deaths).  Not all patients who died were intubated.

The authors also did some sensitivity analyses to try to determine whether the famotidine benefit was really due to some other characteristic or type of care received by those patients.  It’s always helpful to think of alternative reasons why you might have observed your study results and attempt to test whether these other explanations are plausible.  One analysis involved 784 non-study patients who were admitted during the study period and who tested negative for SARS-CoV-2.  Famotidine users were compared to non-users (same <24-hour criterion) and famotidine did not confer any benefit to these COVID-19-negative patients.  So the authors concluded there aren’t other aspects of the famotidine patients that make them “healthier” in general.

They also compared proton-pump inhibitor (PPI) users to non-PPI users to see if the benefit of famotidine was a general benefit due to stomach acid suppression.  In their cohort PPI-users had a higher mortality/intubation rate with an adjusted hazard ratio of 1.34 (95% confidence interval, 1.06-1.09).  This brings up the question of what the expected intubation/mortality rate would be in a population admitted to the hospital but not needing mechanical ventilation for the first 48 hours?  This would help us determine whether the famotidine patients benefited from their drug, or, since PPI-users were part of the non-exposed group, do PPI drugs increase the chances of intubation or death?

In their conclusion the authors state that early administration of famotidine may be important in getting the most benefit out of the drug in treating COVID-19, and that other studies could examine starting it even earlier than what occurred in this study.  Which brings me to that randomized trial, whose results will be known apparently any day now, with its enrollment of critically-ill patients.  It gives the impression that the authors have taken a look at that trial data and decided famotidine doesn’t work in patients with severe disease?

Regarding the inclusion and exclusion criteria, it would have been better to include any patient sick enough to be hospitalized and perhaps not sick enough to need immediate admission to the ICU, without stipulating that the patient couldn’t be intubated or die within 48 hours.  That is the situation that clinicians face and make their decisions about the patient’s care (i.e., without the unknowable 48-hour requirement).

It also would have been useful to collect data on the type and duration of symptoms with which these patients presented to the hospital.  It could either help to identify which patients are most likely to benefit from famotidine or to explain why there was a false association between famotidine use and better outcome.  It also could have facilitated creating equivalent groups of patients, i.e., both sets of patients presenting with similar symptoms and similar duration of symptoms prior to hospitalization.  This is a hole in the study (although sometimes the documentation in the medical record doesn’t allow for this).

Last, I find Dr. Callahan’s use of the phrase “elderly peasants” intriguing.  Does he mean these are poor patients from rural areas?  If that’s the case, are these people more likely to be exposed to other beta coronaviruses through their interactions with animals, and thus receive some partial immunity through that mechanism?  That would make famotidine a true confounder.

In summary, there are some issues with this paper that don’t allow it to provide strong evidence of the effectiveness of famotidine in treating COVID-19.  The results of the randomized trial will be instructive.