What is the current status of phage therapy?

Clinical research on phage therapy is somewhere between comatose and dead.

A search through ClinicalTrials.gov reveals only a few sporadic attempts at bringing PT to the clinic. Many of these are Phase 1 safety/dosing trials which have no apparent follow-up:

• A 2008 P1 trial of PT for venous leg ulcers found no adverse events, but no difference in outcomes for test and control groups [1]
• A 2009 PI/II trial of PT for chronic Pseudomonas ear infections found decent evidence of efficacy [2] . The sponsor of this trial, BioContol Ltd, was absorbed by what is now AmpliPhi Bio, and there is no mention of this project in the Pipeline section of their website. I assume this project is abandoned.
• Pherecydes Pharma ran a PI/II trial of PT on burn wounds in 2015-2016 but is being coy about reporting the results.
• Nestle sponsored a well-designed trial of oral PT for E. coli diarrhea in children and found no evidence of a therapeutic benefit [3] .
• The Phage Therapy Unit of Wroclaw continues to crank out a series of poorly controlled and anecdotal studies [4] that I find utterly unconvincing.
• The Eliava Institute has been using PT for decades in the Republic of Georgia. They have finally gotten around to designing what looks like a decent clinical protocol [5] , but no results are available.

In addition to these trials, a few case and Expanded Access studies have been published:

• AmpliPhi has a nice summary of their EA studies in which patients who had failed antibiotic therapy for S. aureus or Pseudomonas infections were treated. There were a total of 7 patients treated and 6 recovered. I’ll get back to these numbers in a moment.
• Chip Schooley and colleagues published a case study of a patient with a systemic Acinetobacter infection [6] . The patient, who had failed antibiotic therapy, was comatose but stabilized and then recovered after PT.

These last two studies, although reporting clinical success, illustrate two of the critical problems facing phage therapy: market size, and innumeracy on the part of practitioners.

Despite all the hype about the impending superbug apocalypse, the fact is that antibiotic therapy still works nearly all the time. Even in this age of unprecedented levels of resistance, I would guess that first-line antibiotic therapy is effective in about 70% of cases. Second-line therapy works for another 25% or so, and third-line therapy for the remainder. Exceptions are rare enough that they are often published as case studies. Phage therapy would be a fourth-line therapy.

The market for cases that have failed conventional antibiotic therapy is maybe 1000 – 10,000 per year in the US, at least for now. Antibiotics don’t have much pricing power – $2–3k per course of treatment is about all that newer, premium antibiotics like Cubicin and Dificid are able to command.

Let’s say you could get $10k for PT. Even at the upper end of the addressable market, we’re talking about $10M revenue US/year. That’s not a number that is going to elicit excitement from any investor.

This math sucks. It’s why I think we need to socialize development of antibiotics.

More math (TL;DR):

One of the knocks against PT in the pre-antibiotic era was its inconsistency. Given the dosing regimen described in the AmpliPhi and Schooley reports, we still are not addressing one of the likely drivers of inconsistency – the pharmacokinetics of phage.

Phage PK is not going to be at all like that of other drugs, because phage are so much larger and their diffusion rates are correspondingly lower. Small molecule drugs have association constants on the order of 1e7 – 1e8 M-1 min-1; macromolecules like monoclonal antibodies are an order of magnitude slower, and viruses like phage are another order of magnitude slower than that. Small molecules like antibiotics will begin attacking bacteria almost instantly. Phage may take hours.

Most phage scientists don’t appreciate this, because they work in test tubes where high concentrations of phage can be readily achieved, and the rate-limiting factor in phage growth is not initial binding (which is diffusion-dependent) but subsequent synthesis of progeny phage.

But when you inject what seems like a lot of phage (1e9 in both studies) into a human body, two things happen that render test-tube results irrelevant: dilution and clearance.

Phage administered IV are likely confined to the plasma or perhaps whole blood compartments of the body, which have volumes of 3–5L. That 1e9/ml of phage that is injected has (at best) a concentration of 1e5/ml in  the bloodstream.

That’s way too low for the phage to efficiently attack target bacteria.  The typical observed rate of phage infection of bacteria is 1e-9 ml/min in conditions where phage outnumber bacteria. So if phage in the bloodstream are at 1e5/ml, the observed rate of infection is going to be 1e-4 /min. In other words, it takes hours before any appreciable number of phage infections of host bacteria occur.

It gets worse. Phage, like other virus-sized particles, are quickly cleared by Kupffer cells in the liver[7] . Schooley et al, to their credit, measured phage titers post-administration, and rapid clearance is exactly what they found. I replotted their data here and fitted it to a standard two-compartment model of clearance from a bolus dose:

Extrapolating back to t=0 gives us an initial concentration of 2.6e5 phage/ml, implying a volume of distribution of 3.8L, right in line with expectations. The elimination rate of 0.62/min corresponds to a phage half-life in circulation of 1.1 min.

So phage concentration, which was way too low initially, drops 10-fold every 3-4 minutes.

These back of the envelope calculations treat the body like a well-mixed beaker, which is a gross oversimplification. But they give an idea of the challenge. Schooley et al seem to be completely unaware of the problem. They state that the phage dose was limited by endotoxin concentration, and voice no concerns about the expected slow kinetics of phage infection.

I will speculate that the inconsistency of phage therapy outcomes in the literature is due to the “jackpot” aspect of phage infection – if a phage does infect a cell that is part of a biofilm or colony, its progeny phage will be present at a very high local concentration and will infect neighboring bacteria with high efficiency. So we end up with cases where the right phage are given at a too-low dose and often fail, but occasionally find their target and initiate a chain-reaction of bacterial eradication that yields very exciting and promising results.

Until the next time, when it fails again. We seem to be stuck in a kind of Groundhog Day scenario, where no one is addressing the actual challenges of phage therapy, but keep trying the same things that have not been working for a century.

Footnotes

[1] Bacteriophage therapy of venous leg ulcers in humans: results of a phase I safety trial.

[2] A controlled clinical trial of a therapeutic bacteriophage preparation in chronic otitis due to antibiotic-resistant Pseudomonas aeruginosa; a prel… – PubMed – NCBI

[3] Oral Phage Therapy of Acute Bacterial Diarrhea With Two Coliphage Preparations: A Randomized Trial in Children From Bangladesh

[4] Clinical aspects of phage therapy.

[5] Bacteriophages for treating urinary tract infections in patients undergoing transurethral resection of the prostate: a randomized, placebo-controll… – PubMed – NCBI

[6] Development and Use of Personalized Bacteriophage-Based Therapeutic Cocktails To Treat a Patient with a Disseminated Resistant Acinetobacter bauman… – PubMed – NCBI

[7] Rapid and Efficient Clearance of Blood-borne Virus by Liver Sinusoidal Endothelium