Viruses As Medicine: The Return Of Phage Therapy

ATreating a bacterial infection with a virus is an older idea than treating it with a drug. Bacteriophages, viruses that attack bacteria and nothing else, were being dripped onto wounds in Tbilisi and Paris in the 1920s, two decades before penicillin reached the wards, and clinics in the former Soviet Union never entirely gave them up. In the West the practice was pushed aside, and it has come back only because the drugs that pushed it aside are failing. The return is an awkward one. The machinery that licenses a medicine was built around a stable molecule given at a fixed dose, whose behaviour in the body can be set out in advance and repeated from one patient to the next. A phage is nothing of the sort: it is a self-replicating organism that multiplies where its host is abundant, vanishes where it is not, and can be shut out by bacterial resistance within days. Whether the striking recoveries reported in the literature record a treatment effect of the kind trials were designed to detect, or something else that merely resembles one, is now the central dispute.

BFarida Adeyinka, a clinical microbiologist at the Lund Institute of Anti-Infectives, argues that the something else is a filter operating on what reaches print. Between 2016 and 2022 her group assembled a register of 214 compassionate-use cases, in which phages were given to patients whose infections had defeated every licensed antibiotic. In 78 per cent of them the treating team recorded an improvement. Only 31 of the 214, however, included a bacterial count taken both before and after the phages were given, and in 145 cases antibiotics were continued throughout. Her objection is not that the reports are dishonest but that nobody writes up the patient who dies. A clinician who has watched a rescue work submits it to a journal; a clinician who has watched one fail moves on to the next admission. What the register measures, on her account, is the willingness of doctors to describe their successes, and it would produce the same figure whether or not the phages contributed anything.

CThe argument makes a prediction, and Adeyinka has helped test it. If the published benefit is manufactured by selection rather than by the viruses, a study that counts every patient enrolled should find little to report. In 2020 a four-centre trial gave a ready-made preparation of four phages to 96 patients with Pseudomonas infections of diabetic foot ulcers; the wound was cleared in 27 per cent of them, against 41 per cent of those receiving standard care alone. The single encouraging signal was confined to the patients whose own bacteria had been tested against the preparation in the laboratory before treatment began. Adeyinka takes care to present the outcome as a limit on the evidence rather than a verdict on the therapy, and notes that the trial ran for eleven weeks, which may be too short for a slow ulcer to close.

DEwan Lindgren, a molecular ecologist at the Otago Centre for Microbial Ecology, rejects the inference, on the ground that a ready-made preparation tests the convenience of the manufacturer rather than the principle of the treatment. Each phage attacks a narrow range of strains, so four of them, chosen months in advance and grown in bulk, will simply miss most of the bacteria they meet. In his own series, phages were recovered from sewage and matched to each patient's strain within a median of nine days, and 58 per cent of 24 patients cleared the infection. Matching of this kind is now routine in his unit, although it was dismissed as unworkable when he began. Lindgren concedes two things. Every one of his patients received antibiotics as well, so the contribution of the phages cannot be separated from the drugs; and the appearance of resistance was judged from cultures in the laboratory, because no dependable way of following it inside a living patient exists.

EThe quarrel has consequences outside the laboratory. Rules introduced in 2018 in several European jurisdictions allow a phage preparation to be compounded for one named patient without the trial data a marketed product requires, and hospitals have treated cases they would once have had to refuse; a company seeking a licence for an off-the-shelf product, by contrast, must still win the sort of trial that ready-made preparations keep losing. Adeyinka concedes, awkwardly for her own case, that a poor evidence base and an ineffective therapy are two different things, and that her register cannot tell them apart. Lindgren answers that the distinction collapses in practice, because the committees that decide what to fund read thin evidence as evidence of thinness. Neither is happy that the subject has grown up on individual rescues rather than on counts of what becomes of everyone who is treated.

FA truce of sorts is taking shape, in which the two accounts are attached to different situations rather than ranked against each other. Where the organism has been identified and a phage can be found to match it, the individualised approach behaves like a plausible treatment, which would explain the case reports; where a product must be made in advance for patients not yet admitted, mismatch and resistance erode whatever effect there is, which is what the trials keep finding. The question neither side has settled is whether bacteria that survive by shutting a phage out pay for it by turning less dangerous in the body, as they reliably do outside it. No method now available can answer that, and until one exists the choice between the two accounts will go on being made on grounds other than evidence.